You are on page 1of 271

CHLORTEST – EU Funded Research Project under 5FP GROWTH Programme

Resistance of concrete to chloride ingress


– From laboratory tests to in-field performance

Guideline for Practical Use of Methods


for Testing the Resistance of Concrete
to Chloride Ingress

Deliverable D23
CONTRACT N°: G6RD-CT-2002-00855
PROJECT N°: GRD1-2002-71808
ACRONYM: CHLORTEST
DURATION: January 2003 – December 2005
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 2 of 14
Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress

PROJECT
COORDINATOR: SP Swedish National Testing and Research Institute (SP) S

PARTNERS: Institute of Construction Sciences “Eduardo Torroja” (IETcc) E


University of Alicante (UoA) E
Chalmers University of Technology (Chalmers) S
Skanska Norge AS (Selmer) NO
Swedish National Road Administration (SNRA) S
Electricité de France (EDF) F
Netherlands Organisation for Applied Scientific Research (TNO) NL
Hochschule Bremen (HSB) D
Slovenian National Building and Civil Engineering Institute (ZAG) SI
Queens University Belfast (QUB) UK
Laboratório Nacional de Engenharia Civil (LNEC) P
Icelandic Building Research Institute (IBRI) IS
National Institute of Applied Science (INSA) F
Laboratoire Central des Ponts et Chaussées (LCPC) F
Valenciana de Cementos, S.A. CEMEX (VCLC) E
Lund Institute of Technology (LTH) S

ACKNOWLEDGEMENT:
The present document is a deliverable of Workpackage 6 – “Conclusions”. All the consortium
members were involved in the work of this part of the project.

This document was prepared by Tang Luping (SP)

FURTHER INFORMATION:
Regarding this document: Regarding the project in general
Dr Tang Luping Dr Tang Luping
SP Swedish National Testing and Research SP Swedish National Testing and Research
Institute Institute
Box 857 Box 857
S-501 15 BORÅS, Sweden S-501 15 BORÅS, Sweden
Tel. +46-33 165138; Fax: +46-33 134516 Tel. +46-33 165138; Fax: +46-33 134516
e-mail: tang.luping@sp.se e-mail: tang.luping@sp.se
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 3 of 14
Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress

TABLE OF CONTENTS
Page

1 BACKGROUND 5

2 BRIEF DESCRIPTION OF TEST METHODS 6


2.1 European method EN 13396 6
2.2 Nordtest method NT BUILD 443 (Immersion test) 6
2.3 Nordtest method NT BUILD 492 (Rapid migration test) 7
2.4 INSA steady state migration test 7
2.5 Multi-regime migration test 8
2.6 Resistivity test 9

3 PROPOSED METHODS FOR STANDARDISATION 9

4 INTERPRETATION OF TEST RESULTS 9


4.1 Results from Immersion test 9
4.2 Results from Rapid migration test 10
4.3 Results from Resistivity test 11

5 ACCEPTANCE CRITERIA 13

REFERENCES 13
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 4 of 14
Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 5 of 14
Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress

1 BACKGROUND
Concrete exposed to chloride environment such as seawater or roads where de-icing salts are
used in very cold weathers may have durability problems because of the chloride-induced
reinforcement corrosion. The premature deterioration of concrete structures is increasingly
demanding methods for better prediction of the distresses and for evaluation of the suitability
of concrete mixes to the desired service life. In order to recommend reliable methods for
testing the resistance of concrete to chloride ingress, the European Commission funded the
research project CHLORTEST, in which 17 partners from 10 European countries
participated.

Chloride ingress into concrete involves complex physical and chemical processes. The
complexity comes at least from three sources:
a) The external environment is not constant. In marine environments the amount of
chlorides in contact with concrete depends on whether the structure is placed fully
submerged or in the tidal zone, or only in contact with marine fog, while in road
environments, the intermittent use of de-icing salts in very cold weathers will make
difficulty in calculating the amount of chlorides sprayed to the structures;
b) The material concrete is constituted by different types of cement and binder, with
different mix proportions, which makes the concrete to be not a single material but
many different ones which in addition evolve in properties with age; and
c) The mechanisms of chloride penetration are not single (simple diffusion) but
combined with convection (absorption), chemical and physical binding, interaction
of other coexisting ions, etc. Changes in temperature, rain and sunshine introduce
variations that should also be taken into account.
Owing to its important role with regard to durability of concrete structures, many methods
have been proposed for testing chloride ingress in concrete, although the above complexity
has up to now hindered to reach a general agreement on a single test method. A collection of
more than ten different test methods is available through an international committee, RILEM
TC-178 TMC [1]. These methods can be categorised into three categories: diffusion tests,
migration tests, and indirect tests based on resistivity or conductivity. Among those many
methods, six of them were evaluated in this project, that is, European method EN 13396
(immersion test for repair products and systems), Nordtest methods NT BUILD 433
(Immersion test) and NT BUILD 492 (Rapid migration test), INSA steady-state migration
test [2], Multi-regime migration test [3], and Resistivity test [4]. After the pre-evaluation [5],
four methods, that is, NT BUILD 433, NT BUILD 492, Multi-regime migration test, and
Resistivity test, were selected for inter-comparison evaluation, in which 15 laboratories
participated in order to produce reliable precision data [6]. Based on the evaluation results,
three methods, that is, NT BUILD 433, NT BUILD 492 and Resistivity test, are
recommended for further European standardisation.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 6 of 14
Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress

2 BRIEF DESCRIPTION OF TEST METHODS

2.1 European method EN 13396


This method was proposed by CEN/TC 104/SC 8, with the scope of testing the resistance to
chloride penetration in repair products and systems for the protection and repair of concrete.
This method gives chloride contents at different depths after different exposure durations, but
does not provide any transport parameter. The method is also time-consuming and takes
about half a year for the full test.

At the start of CHLORTEST project, the available version of this method is prEN
13396:2002, while the latest formal version is EN 13396:2004. The main difference between
prEN 13396:2002 and EN 13396:2004 is that the former specifies the chloride immersion at
40 °C, while the latter at 23 °C. In this project, the former version was used in the evaluation.

Specimens: 6 specimens of diameter ≥ 100 mm and length ≥ 60 mm, with the trowelled
surface as the test surface

Pre-conditioning: Vacuum saturation with demineralised water

Testing: Immersing specimens in the 3% NaCl solution and measuring chloride contents in
two specimens at the depths 0∼2, 4∼6 and 8∼10 mm, after 28 days, 3 months and 6 months

Test duration: about 6 months

Test results: Chloride contents at three depths and three exposure durations

Precision: Average repeatability COV (Coefficient of Variation) 14% and reproducibility


COV 36% according to the pre-evaluation results [5]

2.2 Nordtest method NT BUILD 443 (Immersion test)


This method is based on natural diffusion under a very high concentration gradient. The test
gives values of Dnssd (non-steady state diffusion coefficient) and Cs (surface total chloride
content) by curve-fitting the measured chloride profile to an error-function solution of Fick’s
2nd law. From the values of Dns and Cs, the parameter KCr, called penetration parameter, can
be derived. The test is relatively laborious and takes relatively long time (more than 35 days).

Specimens: 3 specimens of diameter ≥ 75 mm and length ≥ 60 mm, with the cut surface as
the test surface and epoxy coating on all non-exposure surfaces

Pre-conditioning: Natural immersion in the saturated lime-water until a constant weight is


reached

Testing: Immersing specimens in the solution of 165 g NaCl per litre for at least 35 days, and
measuring chloride penetration profiles by grinding the specimen successively from the
exposed surface and titration-analysing total chloride content in each powder
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 7 of 14
Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress

Test duration: at least 35 days

Test results: Chloride penetration profiles and curve-fitted parameters Dnssd and Cs, as well
as the derived parameter KCr

Precision: Average repeatability COV 20%, 18% and 9% for parameters Dnssd, Cs and KCr,
respectively, and reproducibility COV 28%, 22% and 14% for parameters Dnssd, Cs and KCr,
respectively, according to the final evaluation results [6], noting that KCr is proportional to the
square root of Dnssd, implying that the deviation of KCr is theoretically a half of that of Dnssd

2.3 Nordtest method NT BUILD 492 (Rapid migration test)


This method is a non-steady state migration test using an external electrical field for
accelerating chloride penetration. The test gives values of Dnssm (non-steady state migration
coefficient). The test is relatively simple and rapid with the test duration in most cases 24
hours.

Specimens: 3 specimens of diameter 100 mm and thickness 50 mm, with the cut surface as
the test surface

Pre-conditioning: Vacuum saturation with saturated lime-water

Testing: Imposing a 10∼60 V external potential across the specimen with the test surface
exposing in the 10% NaCl solution and the oppose surface in the 0.3 M NaOH solution for a
certain duration (in most cases 24 hours), then splitting the specimen and measuring the
penetration depth of chlorides by using a colourimetric method

Test duration: 6∼96 (in most cases 24) hours depending on the quality of concrete

Test results: Non-steady state migration coefficient Dnssm calculated from the average
penetration depth

Precision: Average repeatability COV 15% and reproducibility COV 24% for parameter
Dnssm, according to the final evaluation results [6]

2.4 INSA steady state migration test


In this steady state migration test the chloride flux is determined by measuring the
concentration changes in the upstream chloride solution. The test gives values of Ds (steady
state migration coefficient). The test is relatively laborious due to many samples for chloride
analysis. The reproducibility of this test method seems not satisfactory.

Specimens: 3 specimens of diameter 100 mm and thickness 20 mm, with the cut surface as
the test surface

Pre-conditioning: Vacuum saturation with demineralised water


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 8 of 14
Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress

Testing: Imposing a 12 V external potential across the specimen with the test surface
exposing in the 1 M NaCl solution (upstream cell) and the oppose surface in the
demineralised water (downstream cell) and measuring the chloride concentration in the
upstream cell at a certain interval until a constant decrease in chloride concentration can be
obtained from the concentration-time curve

Test duration: a few days up to about two weeks depending on the quality of concrete

Test results: Steady state migration coefficient Ds calculated from the constant flux of
chloride ions

Precision: Average repeatability COV 19% and reproducibility COV 87% for parameter Ds,
according to the pre-evaluation results [5]

2.5 Multi-regime migration test


In this test the chloride flux is determined by measuring the conductivity changes in the
downstream solution. The test gives values of Dssm from the flux and Dnssm from the time-lag.
The test is relatively simple due to the indirect measurement of chloride concentration
through the simple conductivity measurement. The reproducibility of this test method seems
not satisfactory.

Specimens: 3 specimens of diameter 100 mm and thickness 20 mm, with the cut surface as
the test surface

Pre-conditioning: Vacuum saturation with demineralised water

Testing: Imposing a 12 V external potential across the specimen with the test surface
exposing in the 1 M NaCl solution (upstream cell) and the oppose surface in the
demineralised water (downstream cell) and measuring the conductivity, which can be
converted to chloride concentration, in the downstream solution at a certain interval until a
constant increase in conductivity can be obtained from the concentration-time curve

Test duration: a few days up to about two weeks depending on the quality of concrete

Test results: Steady state migration coefficient Ds calculated from the slope of the constant
portion of the concentration-time curve (the constant flux) and non-state migration coefficient
Dns calculated from the intersection on the time-axis of the constant portion of the
concentration-time curve (time-lag)

Precision: Average repeatability and reproducibility COV 22% and 76%, respectively, for
parameter Ds according to the final evaluation results [6], and average repeatability and
reproducibility COV 24% and 45%, respectively, for parameter Dns according to the pre-
evaluation results [5].
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 9 of 14
Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress

2.6 Resistivity test


The resistivity test is an indirect measurement of the transport property of concrete, because
the electrical resistance of concrete is related to the pore structures and ionic strength in the
pore solution.

Specimens: 3 specimens of diameter 100 mm and thickness 50 mm, with the cut surface as
the test surface

Pre-conditioning: Vacuum saturation with distilled or demineralised water.

Testing: Imposing a constant alternative current across the specimen and measuring the
potential response for calculating the resistance using Ohm’s law

Test duration: a few seconds or minutes for the measurements

Test results: Resistivity ρ

Precision: Average repeatability COV 11% and reproducibility COV 25% for resistivity
according to the final evaluation results [6]

3 PROPOSED METHODS FOR STANDARDISATION


Based on the evaluation results, the CHLORTEST consortium proposes the following three
methods for further standardisation at the European level:

• Immersion test (based on NT BUILD 443) for determination of non-steady state


diffusion coefficient, Dnssd, and surface total chloride content, Cs;
• Rapid migration test (based on NT BUILD 492) for determination of non-steady state
migration coefficient, Dnssm, under the standardised laboratory exposure condition;
and
• Resistivity test (based on the version used in [6]) for determination of resistivity ρ as
an indirect measurement of the transport property of concrete
All the above three proposed methods have the precision in an acceptable range, that is,
repeatability COV ≤20% (11%∼20%) and reproducibility COV ≤30% (24%∼28%).
Therefore, they are suitable for data exchanges and industrial applications.

The descriptions of the above proposed methods with certain revisions and modifications for
transfer to standards are given in [7].

4 INTERPRETATION OF TEST RESULTS

4.1 Results from Immersion test


The immersion test provides coupled values of Dnssd and Cs by curve-fitting the measured
chloride profile to an error-function solution of Fick’s 2nd law, which is under the assumption
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 10 of 14
Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress

of constant chloride binding capacity. In the reality, chloride binding capacity is non-linearly
dependent on free chloride concentration and also dependent on type of cementitious binder
[8-10]. The total chloride content is a sum of the free chlorides in the pore solution and the
bound chlorides on the surfaces of hydrates. Therefore, the Cs value is dependent on the
porosity and the type of binder. Even under the same exposure condition, that is, in the
solution of the same chloride concentration, different types of binder will give different
values of Cs.
Since Dnssd is coupled with Cs in the curve-fitting, the value of Dnssd alone may not reflect the
actual resistance of concrete to chloride ingress. To properly interpret the test results, both
values of Dnssd and Cs should be taken into account. The penetration parameter, KCr, combines
influences of Dnssd and Cs and, therefore, better facilitates comparison of the results. An
example is shown in Figure 1, where Mix A reveals a lower value of Dnssd and a higher value
of Cs than Mix B, but both mixes have the same value of KCr.
It should be noted that the parameter KCr with a dimension of mm/√yr is mainly for
facilitating comparison, but not necessarily means the actual penetration depth per square
root of year.

1.4
Mix A Mix B
1.2 -12 2
D nssd = 2 x10 m /s < D nssd = 2.5 x10-12 m2/s
Cl, mass% of concrete

C s = 1.2% of concrete > C s = 0.7% of concrete


1
K Cr = 23 mm/ √yr = K Cr = 23 mm/ √yr
0.8

0.6

0.4

0.2

0
0 2 4 6 8 10

x, mm

Figure 1 – Example of coupled values of Dnssd and Cs.

4.2 Results from Rapid migration test


The rapid migration test provides value of Dnssm, which is also under the assumption of
constant chloride binding capacity during the test. Different from the immersion test, this
assumption may better hold in the rapid migration test, owing to the strong external electrical
field and short testing duration, both of which tend to reduce the amount of bound, especially
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 11 of 14
Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress

physically bound, chlorides. Therefore, the parameter Dnssm describes the property of chloride
transport under a condition of reduced chloride binding [11].

Since Dnssm and Dnssd are from completely different testing conditions, their values may not
be necessarily comparable. However, experimental results from the CHLORTEST project
[5,6] and some other previous projects [12,13] show that these two diffusion coefficients are
coincidentally quite comparable, as shown in Figure 2. Considering the measurement
uncertainties of the test methods, it is reasonable to conclude that both the test methods
measure the similar transport parameters.

40
Ref [5]
Ref [6]
Ref [12]
30
m /s
2

Ref [13]
-12

x ± dx*
D nssm , x10

20 y ± dy*

* dx, dy: reproducibility


10 standard deviation of
D nssd and D nssm ,
respectively.

0
0 10 20 30 40
-12 2
D nssd, x10 m /s

Note: The values of Dnssd were from the exposure for a target of 35 days.

Figure 2 – Relationship between chloride transport parameters Dnssm and Dnssd.

4.3 Results from Resistivity test


Theoretically, resistivity is inversely proportional to diffusivity. Practically, however, the
measured resistivity is contributed by all ions, especially hydroxides, in the pore solution. A
concrete with the binder containing high alkali will resulted in a low resistivity, while a
concrete with pozzolanic additions containing low alkali will often resulted in a high
resistivity. To convert resistivity to chloride diffusion coefficient, the chloride transference
number (the ratio of chloride ions’ intensity to the intensity of total ions) in the pore solution
should be known. It is not an easy task to estimate the intensity of total ions in the pore
solution. Therefore, the resistivity test can only be taken as an indirect measurement of
chloride transport property. Owing to its rapidity and simplicity, this test is a very efficient
tool for quality control in the real production of concrete. Calibration is needed in order to
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 12 of 14
Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress

establish the empiric relationships between resistivity and chloride diffusion coefficient, as
exampled in Figure 3.

5
D = k /ρ Data from [5]
k = 0.213
4
m /s

k = 0.152
Data from [6]
2

3
-12

k = 0.12
D s , x10

Acc. to [14]
2

0
0 5 10 15 20 25 30

1/ρ, mS

60
D = k /ρ k = 3.04 Data from [5]
50
m /s

Data from [6]


2

40
-12

Data from [12]


D nssm , x10

30
k = 0.98
Data from [15]
20
k = 0.86
Data from [16]
10

Data excluded
0 from regression
0 5 10 15 20 25 30

1/ρ, mS

Figure 3 – Example of empiric relationships between resistivity and diffusivity.


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 13 of 14
Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress

5 ACCEPTANCE CRITERIA
In normal cases chloride itself does not directly result in any damage of concrete, but induce
corrosion of steel in concrete. The service-life of reinforced concrete structures exposed to
chloride environments includes the periods of corrosion initiation and propagation. The
former is related to chloride ingress, while the latter is related to corrosion rate. The period of
corrosion initiation is a function of chloride transport property, concrete cover, threshold
chloride level, exposure environment, etc. Obviously, chloride transport property is only one
of the several parameters regarding initiation of corrosion. A concrete with a relatively high
diffusivity can be compensated with a thicker cover to reach the desirable resistance to
corrosion initiation. Therefore, acceptance criteria for the values from the proposed tests are
dependent on many factors, and can be expressed by the following function:

D = f ( x min , t L , C Cr , K env )

where D is the value from a proposed test, f denotes a function, xmin is the minimum
thickness of concrete cover, tL is the desired period of corrosion initiation, CCr is the
threshold chloride level, Kenv is the environmental factor including chloride load (e.g. surface
concentration or content) and micro climate (temperature, humidity, precipitation, etc.).

To solve the above function, proper models are needed. Different prediction models for
chloride ingress have been evaluated in the CHLORTEST project with the infield data
collected from short time (0.5 year) up to 42 years’ exposures [17]. The results show that one
model, Model 5 (ClinConc [18,19]), which was previously calibrated with the 10 years’
traceable data from a field exposure site, reveal reasonably good benchmarks [17]. This
indicates the importance of calibration with the reliable long-term data. When compared with
over 100 years’ service life, however, these 10 years’ traceable data are still not enough to
assure the verification of models for actual service life prediction. Therefore, it is always the
user’s responsibility to use the D values obtained from the proposed tests for modelling of
service life of a particular concrete structure.

REFERENCES
[1] RILEM TC-178 TMC: “Testing and Modelling Chloride Ingress into Concrete”,
Internal document to be published in the near future.
[2] Truc, O., Ollivier, J.P.and Carcassès, M, “A new way for determining the chloride
diffusion coefficient in concrete from steady state migration test”, Cem. Concr. Res.,
30(2) 217-226 (2000).
[3] Castellote M., Andrade C.and Alonso C., “Measurement of the steady and nonsteady
state chloride diffusion coefficients in a migration test by means of monitoring the
conductivity in the anolyte chamber - Comparison with natural diffusion tests”, Cem.
Concr. Res., 31(10) 1411-1420 (2001).
[4] Andrade, C., “Determination of electrical resistivity in concrete specimens – Direct
method”, A translation of UNE 83XXX, 2004.
[5] CHLORTEST, “Resistance of concrete to chloride ingress – from laboratory tests to in-
field performance”, EU-Project (5th FP GROWTH) G6RD-CT-2002-00855, WP2
Report: “Pre-evaluation of different test methods”, 2005.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 14 of 14
Guidelines for practical use of methods for testing the resistance of concrete to chloride ingress

[6] CHLORTEST, “Resistance of concrete to chloride ingress – from laboratory tests to in-
field performance”, EU-Project (5th FP GROWTH) G6RD-CT-2002-00855, WP5
Report: “Final evaluation of test methods”, 2005.
[7] CHLORTEST, “Resistance of concrete to chloride ingress – from laboratory tests to in-
field performance”, EU-Project (5th FP GROWTH) G6RD-CT-2002-00855,
Deliverable 22: “Testing Resistance of Concrete to Chloride Ingress – A proposal to
CEN for consideration as EN standard”, 2005.
[8] Tritthart, J. “Chloride binding in cement: II. The influence of the hydroxide
concentration in the pore solution of hardened cement paste on chloride binding”, Cem.
Concr. Res., 19(5) 683-691 (1989).
[9] Byfors K., “Chloride-initiated Reinforcement Corrosion - Chloride binding”, Swedish
Cement and Concrete Research Institute (CBI), Stockholm, CBI Report 1:90, 1990.
[10] Tang, L. and Nilsson, L-O. “Chloride binding capacity and binding isotherms of OPC
pastes and mortars”, Cem. Concr. Res., 23(2) 347-353 (1993).
[11] Tang, L., “Chloride Transport in Concrete - Measurement and prediction”, Doctoral
thesis, Publication P-96:6, Dept. of Building Materials, Chalmers Universities of
Technology, Gothenburg, Sweden, 1996.
[12] Frederiksen, J.M., Sørensen, H.E., Andersen, A. & Klinghoffer, O., ‘HETEK, The
effect of the w/c ratio on chloride transport into concrete - Immersion, migration and
resistivity tests’, HETEK Report No. 54, ed. by J.M. Frederiksen, published by the
Danish Road Directorate, 1997.
[13] Tang L. and Sørensen, H.E., “Precision of the Nordic Test Methods for Measuring the
Chloride Diffusion/Migration Coefficients of Concrete”, Materials and Structures, 34
479-485 (2001).
[14] Andrade, C., Alonso, C., Arteaga, A. & Tanner, P., “Methodology based on the
electrical resistivity for the calculation of reinforcement service life”. Supplementary
papers of the proceedings of the Fifth International CANMET/ACI Conference on
Durability of concrete. Barcelona, Spain, 4-9 June 2000, pp 899-915.
[15] DURACRETE, “Probabilistic Performance Based Durability Design of Concrete
Structures”, EU Brite-EuRam III project DuraCrete (BE95-1347), Deliverable R8,
“Compliance testing for probabilistic design purposes”, 1999.
[16] Romer, M., “TC 189-NEC Comparative Test - Part I - Comparative test of
'penetrability' methods”, Materials and Structures, 38, 895-905 (2005).
[17] CHLORTEST, “Resistance of concrete to chloride ingress – from laboratory tests to in-
field performance”, EU-Project (5th FP GROWTH) G6RD-CT-2002-00855,
Workpackage 4 Report: “Modelling of chloride ingress”, 2005.
[18] Tang, L. “Engineering Expression of the ClinConc model for prediction of free and
total chloride ingress in submerged marine concrete”, submitted to Cem. Concr. Res.,
2005.
[19] Tang, L., “Service-life prediction based on the rapid migration test and the ClinConc
model”, Proceedings of International RILEM Workshop on Performance Based
Evaluation and Indicators for Concrete Durability, 19-21 March 2006, Madrid, Spain.
CHLORTEST – EU Funded Research Project under 5FP GROWTH Programme

Resistance of concrete to chloride ingress


– From laboratory tests to in-field performance

Testing Resistance of Concrete to Chloride Ingress


– A proposal to CEN for consideration as EN standard

Deliverable D22
CONTRACT N°: G6RD-CT-2002-00855
PROJECT N°: GRD1-2002-71808
ACRONYM: CHLORTEST
DURATION: January 2003 – December 2005
PROJECT
COORDINATOR: SP Swedish National Testing and Research Institute (SP) S

PARTNERS: Institute of Construction Sciences “Eduardo Torroja” (IETcc) E


University of Alicante (UoA) E
Chalmers University of Technology (Chalmers) S
Selmer Skanska AS (Selmer) NO
Swedish National Road Administration (SNRA) S
Electricité de France (EDF) F
Netherlands Organisation for Applied Scientific Research (TNO) NL
Hochschule Bremen (HSB) D
Slovenian National Building and Civil Engineering Institute (ZAG) SI
Queens University Belfast (QUB) UK
Laboratório Nacional de Engenharia Civil (LNEC) P
Icelandic Building Research Institute (IBRI) IS
National Institute of Applied Science (INSA) F
Laboratoire Central des Ponts et Chaussées (LCPC) F
Valenciana de Cementos, S.A. CEMEX (VCLC) E
Lund Institute of Technology (LTH) S

ACKNOWLEDGEMENT:
The present document is a deliverable of Workpackage 6 – “Conclusions”. All the consortium members
were involved in the work of this part of the project. Comments and suggestions from Jens M.
Frederiksen at Birch & Krogboe A/S, Denmark, and Joost Gulikers at RWS, The Netherlands, are
specially appreciated.

This document was prepared by Tang Luping (SP)

FURTHER INFORMATION:
Regarding this document: Regarding the project in general
Dr Tang Luping Dr Tang Luping
SP Swedish National Testing and Research Institute SP Swedish National Testing and Research Institute
Box 857 Box 857
S-501 15 BORÅS, Sweden S-501 15 BORÅS, Sweden
Tel. +46-33 165138; Fax: +46-33 134516 Tel. +46-33 165138; Fax: +46-33 134516
e-mail: tang.luping@sp.se e-mail: tang.luping@sp.se
A proposal to CEN for consideration as EN standard

Contents

Foreword......................................................................................................................................................................5
1 Scope ..............................................................................................................................................................5
2 Normative references ....................................................................................................................................5
3 Terms and definitions ...................................................................................................................................5
4 Test specimens in general............................................................................................................................6
5 Immersion test ...............................................................................................................................................6
5.1 Principle..........................................................................................................................................................6
5.2 Reagents and equipment ..............................................................................................................................7
5.3 Preparation of the test specimen.................................................................................................................7
5.4 Test procedures.............................................................................................................................................8
5.5 Expression of results ....................................................................................................................................9
5.6 Test report ....................................................................................................................................................10
6 Rapid migration test ....................................................................................................................................11
6.1 Principle........................................................................................................................................................11
6.2 Reagents and equipment ............................................................................................................................11
6.3 Preparation of the test specimen...............................................................................................................13
6.4 Test procedures...........................................................................................................................................14
6.5 Expression of results ..................................................................................................................................16
6.6 Test report ....................................................................................................................................................17
7 Resistivity test .............................................................................................................................................18
7.1 Principle........................................................................................................................................................18
7.2 Equipment ....................................................................................................................................................18
7.3 Preparation of the test specimen...............................................................................................................18
7.4 Test procedures...........................................................................................................................................19
7.5 Expression of results ..................................................................................................................................19
7.6 Test report ....................................................................................................................................................20
8 Precision data ..............................................................................................................................................20
9 Annex ............................................................................................................................................................21

Figure 1 — Example of regression analysis for curve-fitting ......................................................................................10

Figure 2 — One arrangement of the rapid migration set-up.......................................................................................12

Figure 3 — Stainless steel clip ...................................................................................................................................12

Figure 4 — Plastic support and cathode ....................................................................................................................13

Figure 5 — Rubber tube assembled with specimen, clips and anode .......................................................................13

Figure 6 — Illustration of measurement for chloride penetration depths ...................................................................16

Figure 7 — Arrangement for the measurement of resistivity by the direct method ....................................................19

Tables

Table 1 — Recommended depth intervals (in mm) for powder grinding......................................................................9

Table 2 — Test voltage and duration for concrete specimen with normal binder content. ........................................15

Table 3 — Precision data of various test methods .....................................................................................................20


3
EU-Project CHLORTEST G6RD-CT-2002-00855
Testing resistance of concrete to chloride ingress

4
A proposal to CEN for consideration as EN standard

Foreword
This is a proposal CEN for consideration as EN standard.

This document is based on the results from the EU-project “Chlortest” under the 5th Frame Programme (GRD1-
2002-71808/G6RD-CT-2002-00855).

Introduction
Chlorides can induce reinforcement corrosion. Reinforced concrete structures exposed to environments containing
chlorides need to be durable and to have an adequate resistance to the ingress of chlorides.

Owing to its important role with regard to durability of concrete structures, many different test methods have been
developed. Some of the commonly used test methods were evaluated in the recently closed EU-project “Chlortest”
under the 5th Frame Programme (GRD1-2002-71808/G6RD-CT-2002-00855). Based on the evaluation results
from this EU-project, the ChlorTest consortium recommends three test methods for testing resistance of concrete to
chloride ingress, that is, immersion test, rapid migration test and resistivity test.

It is desirable, especially in the case of new constituents or new concrete compositions, to test the resistance of
concrete to chloride ingress. This also applies to concrete mixes, concrete products, precast concrete, concrete
members or concrete in situ.

1 Scope
This European prestandard describes the testing for the resistance of concrete to chloride ingress. It can be used
either to supply quantitative data to durability design or to assess the quality of concrete during the production or
construction.

2 Normative references
This proposal incorporates by dated or undated reference, provisions from other publications. These normative
references are cited at the appropriate places in the text and the publications are listed hereafter. For dated
references, subsequent amendments to or revisions of any of these publications apply to this proposal only when
incorporated in it by amendment or revision. For undated references the latest edition of the publication referred to
applies.

EN 12390-2, Testing hardened concrete — Part 2: Making and curing specimens for strength tests.

prEN 14629, Production and systems for the protection and repair of concrete structures – Test methods –
Determination of chloride content in hardened concrete.

ISO 5725, Accuracy (trueness and precision) of measurement methods and results.

3 Terms and definitions


For the purposes of this proposal, the following terms and definitions apply.

3.1
As-cast surface
The surface of a concrete structure exposed to the chloride environment

5
EU-Project CHLORTEST G6RD-CT-2002-00855
Testing resistance of concrete to chloride ingress

3.2
chloride penetration
The ingress of chlorides into concrete due to exposure from external chloride sources

3.3
diffusion
The movement of molecules or ions under a concentration gradient or, more strictly, chemical potential, from a high
concentration zone to a low concentration zone.

3.4
maturity-day
A concrete of 28 maturity-days has developed a maturity corresponding to curing in 28 days at 20 °C.

3.5
migration
The movement of ions under the action of an external electrical field

3.6
profile grinding
Grinding off concrete powder in thin successive layers from a test specimen using a dry process

3.7
resistivity
The electrical resistance per unit length and per unit reciprocal cross-sectional area of concrete at a specified
temperature

3.8
surface-dry condition
A surface condition achieved by drying the water-saturated test specimen with a clean cloth or similar leaving the
test specimen damp but not wet.

4 Test specimens in general


Drilled cores or cast cylinders can be used as test specimens. They must be representative of the concrete and/or
structure in question. The diameter of test specimens is in normal cases 100 mm, or at least 4 times as large as the
maximum size of aggregates. At least three test specimens should be used in the test. If cast cylinders or cores
drilled from cast cube are used as specimens, the casting and curing procedures should be in accordance with EN
12390-2.

In normal cases the concrete should be hardened to at least 28 maturity-days for testing. Since the concrete age
has significant effect on chloride transport, the date of manufacture of concrete and the date of testing shall always
be noted in the report. If the concrete temperature during hardening was outside the range of 10~30 ºC, this must
also be noted in the report.

5 Immersion test

5.1 Principle

A water-saturated concrete specimen is on one plane surface exposed to water containing sodium chloride. After a
specified exposure time thin layers are ground off parallel to the exposed face of the specimen and the chloride
content profile is measured. The apparent chloride diffusion coefficient and the chloride content at the exposed
surface are calculated by curve-fitting the measured profile to the error function solution to Fick’s 2nd law. A
penetration parameter combining the influence of diffusion coefficient, surface chloride content, initial chloride
content and a reference chloride content is calculated for facilitating comparison of the test results from different
types of concrete.

6
A proposal to CEN for consideration as EN standard

5.2 Reagents and equipment

5.2.1 Reagents

5.2.1.1 Distilled or demineralised water.

5.2.1.2 Calcium hydroxide: Ca(OH)2, technical quality.

5.2.1.3 Sodium chloride: NaCl, chemical quality.

5.2.1.4 2-component solvent free (chloride-ion and water vapour diffusion-proof) polyurethane or epoxy-based paint
(membrane).

5.2.1.5 Chemicals for chloride analysis as required by the test method employed (see 5.4.6).

5.2.2 Equipment

5.2.2.1 Equipment for cutting specimens such as a water-cooled diamond saw.

5.2.2.2 Balance: with accuracy better than ±0.01g.

5.2.2.3 Thermometer or thermocouple: with readout device capable of reading to ±1 ºC.

5.2.2.4 Plastic container with tight-fitting lid.

5.2.2.5 Slide calliper with a precision of ±0.1 mm.

5.2.2.6 Ruler with a minimum scale of 1 mm.

5.2.2.7 Equipment for grinding off and collecting concrete powder from thin concrete layers (less than 2mm).

5.2.2.8 Standard sieve, mesh width 1.0 mm.

5.2.2.9 Equipment for chloride analysis as required by the test method employed (see 5.4.6).

5.3 Preparation of the test specimen

5.3.1 Test specimen

5.3.1.1 If drilled cores are used, cut about 70 mm thick slice from the outmost portion of each core as the test
specimen. The surface 10 mm below the as-cast surface is the one to be used for exposure (see 5.3.2.4).

5.3.1.2 If cast cylinders are used, cut an about 70 mm thick slice from the central portion of each cylinder as the test
specimen. The end surface that was nearer to the as-cast surface is the one to be used for exposure (see
5.3.2.4).

5.3.1.3 Cut a slice with a thickness of at least 20 mm from the remainder of the drilled core or cast cylinder for the
measurement of initial chloride content (see 5.4.6.2).

NOTE 1 The term 'cut' here means to saw perpendicularly to the axis of a core or cylinder, using a water-cooled diamond saw.

NOTE 2 It is very important that the test is made on the concrete between the surface and the layer of reinforcement because
it is here the protection against chloride penetration is needed. Furthermore, the quality of the concrete in this
particular area can deviate from the rest of the concrete. The outermost approximately 10 mm of concrete shall be
removed (see 5.3.2.4) to ensure that the measurement is made in an area with relatively constant cement matrix
content and unaffected by surface treatments, lack of curing, etc.

5.3.2 Preconditioning and coating

5.3.2.1 The test specimen is immersed in a saturated Ca(OH)2 solution at 20~25 °C in a tightly closed plastic
container. The container must be filled to the top to minimize carbonation of the liquid. The next day the
mass in surface-dry condition is determined by weighing the test specimen.
7
EU-Project CHLORTEST G6RD-CT-2002-00855
Testing resistance of concrete to chloride ingress

5.3.2.2 The storage in the saturated Ca(OH)2 solution continues until the mass of specimen under the surface-dry
condition does not change by 0.1 mass% when measured at an interval of at least 24 hours.

5.3.2.3 Dry all surfaces of the test specimen except the one to be used for exposure at room temperature to a
stable white-dry condition and given a thick epoxy or polyurethane coating on all white-dry surfaces. It
must be ensured that the method of application and hardening prescribed by the supplier of the coating
material is observed.

5.3.2.4 When the coating has hardened, cut off the outermost approximately 10 mm thick layer from the surface
to be used for exposure and then immerse the test specimen in the Ca(OH)2 solution for overnight.

5.4 Test procedures

5.4.1 Exposure liquid

An aqueous NaCl solution is prepared with a concentration of 165g ± 1g NaCl per litre solution. This exposure
liquid is used for 5 weeks and then replaced by a new pure NaCl solution. The NaCl concentration of the solution
must be checked at least before and after use.

5.4.2 Temperature

The temperature of the water bath shall be kept between 20~25 °C, with a target average temperature of 23 °C
during the immersion. The temperature shall be measured at least once a day.

5.4.3 Exposure

5.4.3.1 Place the surface-dry specimens in the container in such a way that the exposure surface is vertically
exposed.

5.4.3.2 Fill the container with the exposure liquid. It is important that the container is completely filled with the
exposure liquid and closed tightly. The ratio of the exposed area in cm2 to the volume of exposure liquid
in litre shall be between 20~80.

5.4.3.3 Store the container in the room at the temperature as specified in 5.4.2.

5.4.3.4 The exposure shall last for at least 35 days, but in normal cases not exceed 40 days. The date and time
of exposure start and termination is recorded to ±10 minutes.

5.4.4 Profile grinding

5.4.4.1 After the exposure, slightly rinse the specimen with tap water.

5.4.4.2 Wipe off excess water from the surfaces of the specimen.

5.4.4.3 Grind off the material in layers parallel to the exposed surface. The grinding area should be within a
diameter approximately 10 mm less than the full diameter of the core from the exposure surface and can
be gradually reduced in small steps depending on what is appropriate for the grinding tool as the depth
increases, but should always be more than 3 times as large as the maximum size of aggregate. This
obviates the risk of edge effects and disturbances from the coating.

5.4.4.4 At least eight layers must be ground off. The thickness of the layers should be adjusted according to
Table 2 or the expected chloride profile so as to assure minimum 6 points in the profile having chloride
content higher than the initial value (see 5.4.6.2). However, the outermost layer shall always have a
thickness of minimum 1.0 mm.

5.4.4.5 It must be ensured that a sample of at least 5 g of dry concrete dust is obtained from each layer. For each
sample of concrete dust collected, the depth below the exposed surface is calculated as the average of
four uniformly distributed measurements using a slide caliper.

8
A proposal to CEN for consideration as EN standard

Table 1 — Recommended depth intervals (in mm) for powder grinding.

water/binder 0.25 0.30 0.35 0.40 0.50 0.60 0.70


Depth #1 0∼1 0∼1 0∼1 0∼1 0∼1 0∼1 0∼1
#2 1∼2 1∼2 1∼2 1∼3 1∼3 1∼3 1∼5
#3 2∼3 2∼3 2∼3 3∼5 3∼5 3∼6 5∼10
#4 3∼4 3∼4 3∼5 5∼7 5∼8 6∼10 10∼15
#5 4∼5 4∼6 5∼7 7∼10 8∼12 10∼15 15∼20
#6 5∼6 6∼8 7∼9 10∼13 12∼16 15∼20 20∼25
#7 6∼8 8∼10 9∼12 13∼16 16∼20 20∼25 25∼30
#8 8∼10 10∼12 12∼16 16∼20 20∼25 25∼30 30∼35
NOTE For concrete with pozzolanic additions such as fly ash, slag, and silica fume, the depth intervals in the column one place left should
be applied, e.g. for slag cement concrete with w/b = 0.4, the depth intervals in the column for w/b = 0.35 should be applied.

5.4.6 Measurement of chloride content

5.4.6.1 The acid-soluble chloride content of the samples is determined to three decimals in accordance with prEN
14629 or by a similar method with the same or better accuracy.

5.4.6.2 From the concrete slice (see 5.3.1.3), a representative sample of approximately 20 g is prepared by
grinding or crushing until the material passes the 1 mm standard sieve. The initial chloride content is
determined using the same method as used in 5.4.6.1.

5.5 Expression of results

The values of the apparent non-steady state chloride diffusion coefficient Dnssd and the apparent surface chloride
content Cs are determined by curve-fitting the measured chloride profile to equation (1) according to the principle of
least squares, as illustrated in figure 1:

⎛ x ⎞
C(x, t) = C s - ( C s - C i ) erf ⎜ ⎟ (1)
⎜ 4D t ⎟⎠
⎝ nssd

where:

Dnssd: apparent non-steady state diffusion coefficient, m2/s;

x: average depth where the chloride sample was ground, m;

t: immersion duration, seconds;

C(x, t): chloride content at depth x, mass% of sample;

Cs: apparent surface chloride content, mass% of sample;

C i: Initial chloride content, mass% of sample;

erf: error function as defined by equation (2).

2
erf (z ) =
z

π ∫
−u 2
e du (2)
0

The penetration parameter, KCr, is calculated using equation (3), with the above obtained values of Ci, Cs, and Dnssd,
and the assumption of Cr = 0.05 mass% of sample, unless another value is required.

9
EU-Project CHLORTEST G6RD-CT-2002-00855
Testing resistance of concrete to chloride ingress

⎛C -C ⎞
K Cr = 2 D nssd erf -1 ⎜⎜ s r ⎟⎟ (3)
⎝ C s - Ci ⎠

where: erf-1 is the inverse of error function.

NOTE KCr is defined only when Cs > Cr > Ci.

Figure 1 — Example of regression analysis for curve-fitting

5.6 Test report

The test report shall contain at least the following information:

a) Reference to this standard method.

b) Origin, size and marking of the specimens.

c) Concrete identification.

d) Date of manufacture of concrete

e) Any deviation from the test method.

f) Test results, including the specimen dimensions, concrete age at exposure, exposure temperatures, chloride
concentration of the exposure liquid, exposure duration, data of chloride content vs depth, the curve-fitted diffusion
coefficient and surface content, and the calculated penetration parameter.

g) Any observation or suspicion of defects (cracks, large pore holes, foreign bodies etc.) in the specimen.

10
A proposal to CEN for consideration as EN standard

6 Rapid migration test

6.1 Principle

An external electrical field is applied axially across the specimen and forces the chloride ions outside to migrate into
the specimen. After a certain test duration, the specimen is axially split or dry-cut and a silver nitrate solution is
sprayed on to one of the freshly split or cut sections. The chloride penetration depth can then be measured from
the visible white silver chloride precipitation, after which the chloride migration coefficient can be calculated from
this penetration depth.

6.2 Reagents and equipment

6.2.1 Reagents

6.2.1.1 Distilled or demineralised water.

6.2.1.2 Sodium chloride: NaCl, chemical quality.

6.2.1.3 Sodium hydroxide: NaOH, chemical quality.

6.2.1.4 Silver nitrate: AgNO3, chemical quality.

6.2.1.5 Chemicals for chloride analysis as required by the test method employed (optional, see 6.4.6).

6.2.2 Equipment

6.2.2.1 Equipment for cutting specimens such as water-cooled saw.

6.2.2.2 Vacuum container: Capable of containing at least three specimens.

6.2.2.3 Vacuum pump: Capable of maintaining a pressure of less than 50 mbar (5 kPa) in the container, such as
a water-jet pump.

6.2.2.4 Migration set-up: One design is shown in figure 2, which includes the following parts:

a) Silicone rubber tube: inner diameter suitable to the diameter of specimens, thickness 5∼7 mm, long about
150 mm.

b) Clip: stainless steel with 20 mm wideness and diameter range suitable to the rubber tube (figure 3).

c) Catholyte reservoir: plastic box, 370 × 270 × 280 mm (length × width × height).

d) Plastic support: (figure 4).

e) Cathode: stainless steel plate (figure 4), about 0.5 mm thick.

f) Anode: stainless steel mesh or plate with holes (figure 5), about 0.5 mm thick.

NOTE Other designs are acceptable, provided that temperatures of the specimen and solutions during the test can be
maintained in the range of 20 to 25 °C (see 6.4.2).

11
EU-Project CHLORTEST G6RD-CT-2002-00855
Testing resistance of concrete to chloride ingress

+ Potential
- (DC)
a

b
f
c
g
d
h
e

a. Rubber tube e. Catholyte


b. Anolyte f. Cathode
c. Anode g. Plastic support
d. Specimen h. Plastic box

Figure 2 — One arrangement of the rapid migration set-up

6.2.2.5 Power supply: capable of supplying 0~60 V DC regulated voltage with an accuracy of ±0.1 V.

6.2.2.6 Ammeter: capable of displaying current to ±1 mA.

6.2.2.7 Thermometer or thermocouple with readout device capable of reading to ±1 ºC.

6.2.2.8 Any suitable device for splitting (such as compression machine) or dry-cutting (such as air cooled
diamond saw) the specimen.

6.2.2.9 Spray bottle.

6.2.2.10 Slide calliper with a precision of ±0.1 mm.

6.2.2.11 Ruler with a minimum scale of 1 mm.

6.2.2.12 Equipment for chloride analysis as required by the test method employed (optional, see 6.4.6).

Figure 3 — Stainless steel clip

12
A proposal to CEN for consideration as EN standard

Cathode

Plastic
stud

15~20 mm

Plastic
support

32°

Figure 4 — Plastic support and cathode

A-A

Anode* (stainless steel)


Silicone rubber tube

Stainless steel clips A-A

Specimen

* The diameter of anode is 5 mm less than that of the specimen.


Figure 5 — Rubber tube assembled with specimen, clips and anode

6.3 Preparation of the test specimen

6.3.1 Test specimen

6.3.1.1 If drilled cores are used, the outermost approximately 10 mm thick layer should be cut off and the next
(50 ± 2) mm thick slice should be cut from each core as the test specimen. The end surface that was
nearer to the outermost layer is the one to be exposed to the chloride solution (catholyte).

6.3.1.2 If cast cylinders are used, cut a (50 ± 2) mm thick slice from the central portion of each cylinder as the test
specimen. The end surface that was nearer to the as-cast surface is the one to be exposed to the chloride
solution (catholyte).

6.3.1.3 Measure the thickness with a slide calliper and read to 0.1 mm.

NOTE The term 'cut' here means to saw perpendicularly to the axis of a core or cylinder, using a water-cooled diamond saw.

13
EU-Project CHLORTEST G6RD-CT-2002-00855
Testing resistance of concrete to chloride ingress

6.3.2 Preconditioning

After cutting, brush and wash away any burrs from the surfaces of the specimen, and wipe off excess water from
the surfaces of the specimen. When the specimens are surface dry, place them in the vacuum container for
vacuum treatment. Both end surfaces must be exposed. Reduce the absolute pressure in the vacuum container to
a pressure in the range of 10~50 mbar (1~5 kPa) within a few minutes. Maintain the vacuum for three hours and
then, with the vacuum pump still running, fill the container with the distilled or demineralised water so as to immerse
all the specimens. Maintain the vacuum for a further hour before allowing air to re-enter the container. Keep the
specimens in the solution for (18 ± 2) hours.

6.4 Test procedures

6.4.1 Catholyte and anolyte

The catholyte solution is 10% NaCl by mass in tap water (100 g NaCl in 900 g water, about 2 N) and the anolyte
solution is 0.3 M NaOH in distilled or demineralised water (approximately 12 g NaOH in 1 litre water). Store the
solutions at a temperature of 20~25 °C.

NOTE It is important to use distilled or demineralised water for the anolyte solution to obviate the corrosion damage of
anode.

6.4.2 Temperature

Maintain the temperatures of the specimen and solutions in the range of 20~25 °C during the test.

6.4.3 Preparation of the test

6.4.3.1 Fill the catholyte reservoir with about 12 litres of 10 % NaCl solution.

6.4.3.2 Fit the rubber tube on the specimen as shown in figure 4 and secure it with two clips. If the curved surface
of the specimen is not smooth, or there are defects on the curved surface which could result in significant
leakage, apply a line of silicone sealant to improve the tightness.

6.4.3.3 Place the specimen on the plastic support in the catholyte reservoir (see figure 2).

6.4.3.4 Fill the tube above the specimen with about 300 ml anolyte solution (0.3 M NaOH) in order to cover the
specimen surface with at least 3 mm liquid layer.

6.4.3.5 Immerse the anode in the anolyte solution.

6.4.3.6 Connect the cathode to the negative pole and the anode to the positive pole of the power supply.

6.4.4 Migration test

6.4.4.1 Turn on the power, with the voltage preset at 30 V, and record the initial current through each specimen.

6.4.4.2 Adjust the voltage if necessary (as shown in Table 2). After adjustment, note the value of the initial
current again.

6.4.4.3 Record the initial temperature in each anolyte solution, as shown by the thermometer or thermocouple.

6.4.4.4 Choose appropriate test duration according to the initial current (see Table 2).

6.4.4.5 Record the final current and temperature before terminating the test.

14
A proposal to CEN for consideration as EN standard

Table 2 — Test voltage and duration for concrete specimen with normal binder content.

Initial current I30V Applied voltage U Possible new initial current I0 Test duration t (hour)
(with 30 V) (mA) (after adjustment) (V) (mA)
I0 < 5 60 I0 < 10 96 or longer
5 ≤ I0 < 10 60 10 ≤ I0 < 20 48
10 ≤ I0 < 15 60 20 ≤ I0 < 30 24
15 ≤ I0 < 20 50 25 ≤ I0 < 35 24
20 ≤ I0 < 30 40 25 ≤ I0 < 40 24
30 ≤ I0 < 40 35 35 ≤ I0 < 50 24
40 ≤ I0 < 60 30 40 ≤ I0 < 60 24
60 ≤ I0 < 90 25 50 ≤ I0 < 75 24
90 ≤ I0 < 120 20 60 ≤ I0 < 80 24
120 ≤ I0 < 240 15 60 ≤ I0 < 120 24
240 ≤ I0 < 600 10 80 ≤ I0 < 200 24
I0 > 600 10 I0 > 200 6
NOTE 1: The values in the table are based on the specimens with 100 mm diameter. For the specimens with another diameter,
2
d (mm), correct the measured current by multiplying a factor of (100/d) in order to be able to use the above table.

NOTE 2: For specimens with a special binder content, such as repair mortars or grouts, correct the measured current by
multiplying a factor (approximately equal to the ratio of normal binder content to actual binder content) in order to be
able to use the above table.

6.4.5 Measurement of chloride penetration depth

6.4.6.1 Remove the specimen by following the reverse of the procedure in 6.4.3. A wooden rod is often helpful in
removing the rubber tube from the specimen.

6.4.6.2 Rinse the specimen with tap water.

6.4.6.3 Wipe off excess water from the surfaces of the specimen.

6.4.6.4 Split or dry-cut the specimen axially into two pieces. Choose the piece having the split or cut section more
nearly perpendicular to the end surfaces for the penetration depth measurement, and keep the other
piece for chloride content analysis (optional). If there is no requirement for chloride content analysis, both
the pieces can be used for the penetration depth measurement.

6.4.6.5 Spray 0.1 M silver nitrate solution on to the freshly split or cut section.

6.4.6.6 When the white silver chloride precipitation on the split or cut surface is clearly visible (after about 15
minutes), measure the penetration depth, with the help of the slide calliper and a suitable ruler, from the
centre to both edges at intervals of 10 mm (see figure 6) to obtain seven depths. Read the depth to 0.1
mm.

NOTE 1 The term 'dry-cut' here means to saw the specimen using an air-cooled diamond saw.

NOTE 2 If the penetration front to be measured is obviously blocked by the aggregate, move the measurement to the nearest
front where there is no significant blocking by aggregate or, alternatively, ignore this depth if there are more than five
valid depths.

NOTE 3 If there is a significant defect in the specimen that results in a penetration front much larger than the average, ignore
this front as indicative of the penetration depth, but note or photograph it and report the condition.

NOTE 4 To obviate the edge effect due to a non-homogeneous degree of saturation or possible leakage, do not make any
depth measurements in the zone within about 10 mm from the edge (see figure 6).

15
EU-Project CHLORTEST G6RD-CT-2002-00855
Testing resistance of concrete to chloride ingress

10 mm 10 mm
Measurement zone

Specimen

L
xd6 xd4 xd2 xd1 xd3 xd5 xd7

Ruler
10 10 10 10 10 10 mm

Figure 6 — Illustration of measurement for chloride penetration depths

6.4.6 Measurement of surface chloride content (optional)

6.4.6.1 From the other axially split or cut specimen, cut an approximately 5 mm thick slice parallel to the end
surface that was exposed to the chloride solution (catholyte).

6.4.6.2 Determine the chloride content in the slice in accordance with prEN 14629 or by a similar method with the
same or better accuracy.

NOTE 1 Information about chloride binding capacity of the tested material may be estimated from the surface chloride content.

NOTE 2 The thickness of the slice should always be less than the minimum penetration depth.

6.5 Expression of results

Calculate the non-steady state migration coefficient from equation (4):

RT x d − α x d
Dnssm = ⋅ (4)
z FE t

where:

U−2
E= (5)
L

RT ⎛ 2c ⎞
α=2 ⋅ erf −1 ⎜ 1 − d ⎟ (6)
z FE ⎝ c0 ⎠

Dnssm: non-steady state migration coefficient, m2/s;

z: absolute value of ion valence, for chloride, z = 1;

F: Faraday constant, F = 9.648 ×104 J/(V·mol);

U: absolute value of the applied voltage, V;

16
A proposal to CEN for consideration as EN standard

R: gas constant, R = 8.314 J/(K·mol);

T: average value of the initial and final temperatures in the anolyte solution, K;

L: thickness of the specimen, m;

xd: penetration depth, m;

t: test duration, seconds;

erf-1: inverse of error function;

cd: chloride concentration at which the colour changes, cd ≈ 0.07 N for OPC concrete;

c0: chloride concentration in the catholyte solution, c0 ≈ 2 N.

⎛ 2 × 0.07 ⎞
Since erf −1⎜ 1 − ⎟ = 1.28, the following simplified equation can be used:
⎝ 2 ⎠

0.0239( 273 + T ) L ⎛
⎜ x − 0.0238 ( 273 + T ) L xd ⎞⎟
Dnssm = (7)
(U − 2) t ⎜ d
⎝ U−2 ⎟

where:

Dnssm: non-steady-state migration coefficient, ×10-12 m2/s;

U: absolute value of the applied voltage, V;

T: average value of the initial and final temperatures in the anolyte solution, °C;

L: thickness of the specimen, mm;

xd: penetration depth, mm;

t: test duration, hour.

6.6 Test report

The test report shall contain at least the following information:

a) Reference to this standard method.

b) Origin, size and marking of the specimens.

c) Concrete identification.

d) Date of manufacture of concrete

e) Any deviation from the test method.

f) Test results, including the specimen dimensions, concrete age at testing, applied voltage, initial and final currents,
initial and final temperatures, testing duration, average and maximum penetration depth, average and maximum
migration coefficient, and individual data of penetration depths.

g) Any observation or suspicion of an abnormal penetration front due to defects (cracks, large pore holes, foreign
bodies etc.) in the specimen.
h) Optional information about surface chloride content.

17
EU-Project CHLORTEST G6RD-CT-2002-00855
Testing resistance of concrete to chloride ingress

7 Resistivity test

7.1 Principle

The test is based on the measurement of the electrical resistance of a concrete sample by means of current lines
parallel to the basement of the sample.

7.2 Equipment

7.2.1 Equipment for cutting specimens such as a water-cooled diamond saw.

7.2.2 Vacuum container: Capable of containing at least three specimens.

7.2.3 Vacuum pump: Capable of maintaining a pressure of less than 50 mbar (5 kPa) in the container, such as
a water-jet pump.

7.2.4 Resistivimeter: able to apply a stable alternating voltage or current at a frequency between 50∼100Hz to
the test specimen and to measure respectively the current or voltage drop generated.

NOTE It is acceptable to use an alternating current source with two external multimeters to measure the current and voltage
drop.

7.2.5 Electrodes: metallic nets (openings < 2mm) or plates, made of steel, copper or any other good conductive
metal and free of superficial impurities (deposits, rust, oxides, etc.), with a diameter equal to that of the
specimen.

7.2.6 Contact sponges: two pieces of thin sponges (thickness < 5mm) with a diameter equal that of electrodes.

7.2.7 Weight object: made of non-conductive material with a mass of about 2 kg.

7.2.8 Ruler with a minimum scale of 1 mm.

7.3 Preparation of the test specimen

7.3.1 Test specimen

7.3.1.1 If drilled cores are used, the outermost approximately 10 mm thick layer should be cut off and the next
(50 ± 2) mm thick slice should be cut from each core as the test specimen.

7.3.1.2 If cast cylinders are used, cut a (50 ± 2) mm thick slice from the central portion of each cylinder as the test
specimen.

7.3.1.3 Measure the thickness with a slide calliper and read to 0.1 mm.

NOTE The term 'cut' here means to saw perpendicularly to the axis of a core or cylinder, using a water-cooled diamond saw

7.3.2 Preconditioning

After cutting, brush and wash away any burrs from the surfaces of the specimen, and wipe off excess water from
the surfaces of the specimen. When the specimens are surface dry, place them in the vacuum container for
vacuum treatment. Both end surfaces must be exposed. Reduce the absolute pressure in the vacuum container to
a pressure in the range of 10~50 mbar (1~5 kPa) within a few minutes. Maintain the vacuum for three hours and
then, with the vacuum pump still running, fill the container with the distilled or demineralised water so as to immerse
all the specimens. Maintain the vacuum for a further hour before allowing air to re-enter the container. Keep the
specimens in the solution for (18 ± 2) hours.

NOTE If the specimens are in whole time cured in the water and no apparent drying has happened during transport, drilling,
cutting, and other preparation procedures, the vacuum procedure can be skipped.

18
A proposal to CEN for consideration as EN standard

7.4 Test procedures

7.4.1 Testing room

The testing room shall have a temperature of (20 ± 2) ºC and a relative humidity RH ≥ 45%.

7.4.2 Measurement arrangement

The measurement arrangement is as shown in figure 7.

2 kg weight
Electrodes
Resistivimeter
Sponges

Specimen

Holder

Figure 7 — Arrangement for the measurement of resistivity by the direct method

7.4.3 Measurement of the ohmic resistance

7.4.3.1 Wet the sponges with tap water and slightly drip them to remove the excess of water.

7.4.3.2 Wipe away the exceed water from the surfaces of the samples.

7.4.3.3 Place the sponges, specimen, electrodes and the 2 kg weight in such a way as shown in figure 7.

7.4.3.4 Impose a current, I, of about 40 mA from the resistivimeter and record the correspondent potential drop to
1 mV as soon as the value becomes stable (after around 5 seconds) , noted as Us+sp.

7.4.3.5 Turn off the current and remove the specimen but keep the sponges, electrodes and the 2 kg weight in
the measurement arrangement.

7.4.3.6 Impose a current, I, of about 40 mA from the resistivimeter and record the correspondent potential drop to
1 mV as soon as the value becomes stable (after around 5 seconds) , noted as Usp.

NOTE The value of Usp/I should not be higher than 100 mV/mA. Otherwise the sponges shall be rewetted and
the measurement repeated.

7.5 Expression of results

Calculate the resistivity from equation (8):

πd 2 U s + sp − U sp
ρ= ⋅ (8)
4000L I

19
EU-Project CHLORTEST G6RD-CT-2002-00855
Testing resistance of concrete to chloride ingress

where:

ρ: resistivity of concrete, Ω·m;

d: diameter of the specimen, mm;

L: thickness of the specimen, mm;

Us+sp: absolute value of the potential drop recorded with specimen and sponges, mV;

Usp: absolute value of the potential drop recorded with sponges only, mV;

I: Imposed current, mA.

7.6 Test report

The test report shall contain at least the following information:

a) Reference to this standard method.

b) Origin, size and marking of the specimens.

c) Concrete identification.

d) Date of manufacture of concrete

e) Any deviation from the test method.

f) Test results, including the specimen dimensions, concrete age at testing, temperature, imposed current, potential
drops, and resistivity.

g) Any observation or suspicion of defects (cracks, large pore holes, foreign bodies etc.) in the specimen.

8 Precision data
Based on the inter-laboratory test organised in the EU-project “Chlortest” (GRD1-2002-71808/G6RD-CT-2002-
00855) with 3 replicates and 12 laboratories, the values of repeatability and reproducibility of various test methods
are listed in Table 3.

Table 3 — Precision data of various test methods

Test method Parameter Repeatability sr Reproducibility sR


2
Rapid migration test Dnssm sr = 0.152Dnssm (R = 0.86) sR = 0.236Dnssm (R2 = 0.93)

Resistivity test ρ sr = 0.105ρ (R2 = 0.97) sR = 0.251ρ (R2 = 998)

Dnssd sr = 0.201Dnssd (R2 = 0.94) sR = 0.283Dnssd (R2 = 0.99)

Immersion test Cs sr = 0.177Cs (R2 = 0.14) sR = 0.223Cs (R2 = 0.41)

KCr sr = 0.090KCr (R2 = 0.76) sR = 0.135KCr (R2 = 0.88)

Ds sr = 0.220Ds (R2 = 0.92) sR = 0.759Ds (R2 = 0.95)


Steady state migration test
Dns sr = 0.237Dns (R2 = 0.76) sR = 0.454Dns (R2 = 0.92)

20
A proposal to CEN for consideration as EN standard

9 Annex
Major modifications in the proposed methods for standardisation are summarised in the following table.

Proposed Method Method Used in ChlorTest

Immersion Test NT BUILD 443

• Cut off the outmost about 10 mm layer after • Cut off the outmost about 10 mm layer before
lime-water saturation and coating to prevent lime-water saturation and coating
the exposure surface from calcium
densification
• Exposure duration “at least 35 days, but in • Exposure duration “at least 35 days”
normal cases not exceed 40 days” to reduce
the age effect in the normal tests

Rapid Migration Test NT BUILD 492

• Vacuum saturation with distilled or • Vacuum saturation with saturated lime-water


demineralised water to simplify the
procedure, because during this short
saturation period there should not be
remarkable leaching problem
• After the migration test, “split or dry-cutting • After the migration test, “split the specimen”
the specimen”, to facilitate the laboratory that
may have no machine for splitting and also to
make the depth measurement easier
Resistivity test UNE 83XXX (Carmen Andrade’s translation)

• Thickness of specimen is specified to (50 ± • No size specification


2) mm to reduce the size effect
• Vacuum to a pressure of “less than 50 mbar • Vacuum to a pressure of “less than 1 mmHg”*
(5 kPa) in the container”, to be more
practical, especially when specimens are
moist or water has been filled in the container
* Corresponding to 0.76 mbar or 0.076 kPa. This pressure was specified in ASTM C1202, but can hardly be achieved without a
special vacuum pump, especially when specimens are moist or water has been filled in the container.

21
CHLORTEST – EU Funded Research Project under 5FP GROWTH Programme

Resistance of concrete to chloride ingress


– From laboratory tests to in-field performance

WP2 REPORT – PRE-EVALUATION OF


TEST METHODS

Deliverables D5-8
CONTRACT N°: G6RD-CT-2002-00855
PROJECT N°: GRD1-2002-71808
ACRONYM: CHLORTEST
DURATION: January 2003 – December 2005
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 2 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

PROJECT
COORDINATOR: SP Swedish National Testing and Research Institute (SP) S

PARTNERS: Institute of Construction Sciences “Eduardo Torroja” (IETcc) E


University of Alicante (UoA) E
Chalmers University of Technology (Chalmers) S
Skanska Norge AS (Selmer) NO
Swedish National Road Administration (SNRA) S
Electricité de France (EDF) F
Netherlands Organisation for Applied Scientific Research (TNO) NL
Hochschule Bremen (HSB) D
Slovenian National Building and Civil Engineering Institute (ZAG) SI
Queens University Belfast (QUB) UK
Laboratório Nacional de Engenharia Civil (LNEC) P
Icelandic Building Research Institute (IBRI) IS
National Institute of Applied Science (INSA) F
Laboratoire Central des Ponts et Chaussées (LCPC) F
Valenciana de Cementos, S.A. CEMEX (VCLC) E
Lund Institute of Technology (LTH) S

ACKNOWLEDGEMENT:
The present document is deliverables of Workpackage 2 – “Pre-Evaluation of Test Methods”.
The consortium members IETcc, SP, TNO, QUB, LNEC, IBRI, INSA and LCPC were
involved in the work of this part of the project. The work was led by IETcc and assisted by
SP.

This document was prepared by Marta Castellote and Carmen Andrade (IETcc)

FURTHER INFORMATION:
Regarding this document: Regarding the project in general
Dr Carmen Andrade Dr Tang Luping
Consejo Superior de Investigaciones Cientificas SP Swedish National Testing and Research Institute
Serrano Galvache s/n Box 857
ES-28033 MADRID, Spain S-501 15 BORÅS, Sweden
Tel. +34-91 3020440; Fax: +34-91 3020700 Tel. +46-33 165138; Fax: +46-33 134516
e-mail: andrade@ietcc.csic.es e-mail: tang.luping@sp.se
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 3 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

TABLE OF CONTENTS
Page

1 INTRODUCTION 5

2 PARTICIPATING LABORATORIES 5

3 TEST METHODS SELECTED 6

4 MIXES 6

5 TEST RESULTS AND PRECISION ANALYSIS 4

6 STATISTICAL ANALYSIS OF THE DATA 13


7 DISCUSSION 18

8 CONCLUDING REMARKS AND SUGGESTIONS 27

REFERENCES 28
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 4 of 28
WP2 REPORT – Pre-Evaluation of Test Methods
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 5 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

1 INTRODUCTION
According to work-plan of the Chlortest project, the Objectives to be reached by the
Workpackage 2 – Pre-evaluation of different test methods – are:

• Pre-evaluation of five to six promising test methods selected from WP1


• Identify about three best methods for final evaluation in WP5

To reach these objectives, and according to the work plan, six test methods were selected
from WP1 and their results have been evaluated. From the results, four methods have been
identified for final evaluation in WP5 on a basis of their values, accuracy, information that
supply and practical convenience.

2 PARTICIPATING LABORATORIES
The laboratories that have participated in this WP are listed in table 1, where it can be seen that, in
addition to the seven labs involved in WP2 of the project, LCPC was a volunteer laboratory to
perform also the tests, what was accepted from the consortium.

In table 1, the different laboratories have been listed in alphabetical order, not being the number
that appears in the first column of the table valid for identification of the results of each lab. On
this purpose, each of them has been given randomly a different number.

Table 1: Laboratories that took part in the WP2.


Number Laboratory Country
1 IBRI Iceland
2 IETcc Spain
3 INSA France
4 LCPC France
5 LNEC Portugal
6 QUB United Kingdom
7 SP Sweden
8 TNO Netherlands
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 6 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

3 TEST METHODS SELECTED

During the kick- off meeting, the participants agreed upon the following six test methods for
pre-evaluation:

• D2: NT BUILD 443 [1]: Immersion in 165g NaCl per litre at 23 °C for 35d.

• R1: Resistivity test [2,3]: Measuring resistivity of concrete and calculation of the
effective diffusion coefficient.

• M3: INSA steady-state-migration [4],: Measuring concentration changes in the


upstream cell.
• M4: NT BUILD 492 [5]: Non-steady state migration under 10~60 V for 24 h.

• M6: Multi-regime method [6]: Calculation of steady and non-steady state diffusion
coefficients by measuring conductivity change in the downstream cell.

• PrEN: prEN 13396 [7]: Immersion in 3% NaCl at 40 °C for 28, 90 and 180 days

A detailed description of all the selected six methods for pre-evaluation are available at the
references 1 to 7 and at the RILEM TC 178’s homepage.

4 MIXES
It was agreed at the kick-off meeting to cast concrete specimens in different countries with a
typical mixture design used in the respective country. Upon the agreement, Chalmers
(Sweden) prepared the concrete specimens with 5% silica fume and w/b 0.4. The IETcc
(Spain) prepared the OPC (CEM I) concrete specimens with w/c 0.45. LNEC (Portugal)
prepared the concrete specimens with CEM IV (39% fly ash) and w/b 0.45 and TNO
(Netherlands) prepared the concrete specimens with CEM III (76% slag) and w/b 0.45.

These four mixes were cast, cured, drilled and the cores delivered according to the procedures
given in the quality assurance procedure (QAP).

The mixture proportions of the concrete manufactured for WP2 are summarized in Table 2.

5 TEST RESULTS
As has been commented, each laboratory has been labelled with a number that has been
randomly assigned in order to maintain the confidentiality of the results corresponding to
each laboratory.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 7 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

Tables 3 to 7 present the individual results for every test. It can be noticed that most of the
laboratories, according to the recommended procedure, have performed the different tests by
triplicate. Only a few numbers of laboratories have not reported the whole set of results.

Table 2 Mixture proportions of concrete manufactured for WP2.


Kg/m3 SF OPC FA SL
Cast by CHALMERS IETcc LNEC TNO
Cement 399 400 340 350
I-42.5 N I-42.5 R/SR IV/B 32.5 R III/B 42.5 LH
V/SR/LA HS
Silica fume 21 (slurry) ----- -----
Fly ash ----- ----- -----
Slag ----- ----- -----
Water 168 180 153 157.5
Sand 842.5 (0-8 742 (0-6 mm) 62 (0-2 mm) 70 (0-1 mm)
mm) 603 (0-4 mm) 790 (0-4 mm)
Coarse 842.5 (8-16 1030 (6-16 619 (4-12 mm) 1040 (4-16
aggregate mm) mm) 555 (12-25 mm)
mm)
Total aggregate 1685 1772 1823 1830
Super- 3.4 4.8 4.1 3.9
plasticisers Cementa 92M Rheobuild 1000 Cretoplast
Melcret 222
Air content 6% ----- 1.5
W/C 0.42 0.45
W/B 0.40 0.45 0.45 0.45
Strength (MPa) 63 45 52.6
Slump (mm) > 150
Casting date November 2002 May 2003 June 2003 October 2003
Delivered May 2003 August 2003 October 2003 April 2004

Concerning the results for prEN 13396, it has been considered that the increment nº 2 was the
most representative for the analysis, and only these increment, for each time of exposure, has
been taken into account in the statistical treatment the results.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 8 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

Table 3: Individual results for method D2 corresponding to the Dns “effective chloride
transport coefficient” and K, “penetration parameter”.
Method D2 (Dns x 10-12 m2/s) Method D2 (K mm/year1/2)
Laboratory Mix 1 Mix 2 Mix 3 Mix 4 Laboratory Mix 1 Mix 2 Mix 3 Mix 4
Lab 1, test 1 2,01 16,9 9 1,8 Lab 1, test 1 21 62 47 21
test 2 3,73 18,2 5,5 2,5 test 2 28 62 37 24
test 3 2,33 16,4 9,8 2 test 3 23 60 50 22
Lab 2, test 1 2,01 17,94 5,11 1,9 Lab 2, test 1 19,3 58,31 31,34 19,14
test 2 1,61 19,97 5,85 1,08 test 2 17,2 61,25 33,22 14,26
test 3 1,51 14,47 5,13 1,72 test 3 16,6 52,17 31,03 18,2
Lab 3, test 1 1,03 15,01 4,42 4,43 Lab 3, test 1 14,05 52,18 28,7 28,25
test 2 1,12 13,13 4,78 2,88 test 2 14,11 48,75 30,2 22,77
test 3 1,92 18,57 3,5 3,64 test 3 18,47 58,14 26,08 25,65
Lab 4, test 1 1,24 7,46 1,91 Lab 4, test 1 14,97 36,96 18,92
test 2 2,33 4,9 3,06 test 2 20,7 30,21 24,14
test 3 1,26 9,77 4,6 test 3 15,06 42,94 31,42
Lab 5, test 1 0,69 18,19 6,16 2,03 Lab 5, test 1 14 63 38 22
test 2 1,23 32,38 5,76 2,19 test 2 16 81 37 23
test 3 1,13 37,56 14,84 3,26 test 3 17 86 56 24
Lab 6, test 1 4,32 42,7 2,97 Lab 6, test 1 28,26 232 23,39
test 2 3,2 65,4 3,28 test 2 24,34 287 24,91
test 3 3,63 29,2 2,67 test 3 25,88 192 22,46
Lab 7, test 1 1,82 12,7 5,84 2,12 Lab 7, test 1 22 54 38 23
test 2 1,8 19,7 6,11 2,06 test 2 21 64 36 23
test 3 9,64 7,15 2,49 test 3 44 38 25
Lab 8, test 1 2,35 10,03 15,61 Lab 8, test 1 23 44 62
test 2 2,13 12,35 10,88 test 2 23 49 53
test 3 2,44 14,1 8,52 test 3 23 51 43

Table 4: Individual results for methods M3 (INSA steady-state-migration) and M4 (NT


BUILD 492).
Method M3 (Ds x 10-12 m2/s) Method M4 (Dns x 10-12 m2/s)
Laboratory Mix 1 Mix 2 Mix 3 Mix 4 Laboratory Mix 1 Mix 2 Mix 3 Mix 4
Lab 1, test 1 0,62 0,79 Lab 1, test 1 2,6 16,3 7,3 2,4
test 2 0,61 1,1 test 2 2,7 15,9 7,3 3,2
test 3 0,81 0,88 test 3 3 14,1 8,6 2,4
Lab 2, test 1 1,56 1,928 1,63 1,55 Lab 2, test 1 2,2 15,5 8,3 1,8
test 2 1,58 1,524 1,54 1,68 test 2 2,7 17 6,7 2,34
test 3 1,64 1,553 1,71 1,64 test 3 3,4 15,6 7,5 1,68
Lab 3, test 1 0,093 0,2 0,24 0,08 Lab 3, test 1 2,7 15,1 3,8 2,3
test 2 0,136 0,25 0,16 0,06 test 2 2 17 6 2,8
test 3 0,23 0,13 0,1 test 3 2,7 15,3 4,6 3,6
Lab 4, test 1 23,1 64,31 33,63 0,497 Lab 4, test 1 1,8 12,8 4,3 2,5
test 2 82,45 63,68 49,83 0,406 test 2 1,9 18,6 3,5 2
test 3 99,76 63,38 65,02 0,39 test 3 2 14,1 3,3 2,1
Lab 5, test 1 1 5,2 1,5 0,7 Lab 5, test 1 2,8 17,7 3,2 2,2
test 2 0,8 5,8 1,5 0,4 test 2 4,3 17,2 3,2 2,5
test 3 0,8 4,7 2,1 0,9 test 3 3,4 15,7 6 2,3
Lab 6, test 1 1,37 2,8 Lab 6, test 1 1,59 4,4 2,4
test 2 1,53 3,02 test 2 2,9 4,7 81,1
test 3 1,1 2,92 test 3 2 3,1 3,6
Lab 7, test 1 0,31 0,61 0,26 0,2 Lab 7, test 1 3,1 14,9 4,8 2,2
test 2 0,38 0,5 0,31 0,25 test 2 3 14,8 5,7 2,9
test 3 0,33 0,63 0,3 0,29 test 3 3 14,9 5,4 2,9
Lab 8, test 1 1,41 2,39 1,31 Lab 8, test 1 2 15,4 3,4
test 2 1,6 2,83 1,94 test 2 2,6 21,2 6,5
test 3 2,26 2,14 0,97 test 3 2,3 13,6 4
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 9 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

Table 5: Individual results for method M6 (Multi-regime method) corresponding to the Ds


(steady state diffusion coefficient) and Dns (non-steady state diffusion coefficient).
Method M6 (Ds x 10-12 m2/s) Method M6 (Dns x 10-12 m2/s)
Laboratory Mix 1 Mix 2 Mix 3 Mix 4 Laboratory Mix 1 Mix 2 Mix 3 Mix 4
Lab 1, test 1 0,42 1,43 0,37 0,19 Lab 1, test 1 4,8 16,4 17,6 3,9
test 2 0,28 1,73 0,29 0,22 test 2 4,5 15,1 45 3,8
test 3 0,46 1,47 0,36 0,19 test 3 7,6 17,4 10,5 4,1
Lab 2, test 1 1,33 2,1 1,055 0,59 Lab 2, test 1 4,2 13,6 9,27 2,28
test 2 1,88 2,1 0,62 0,66 test 2 4,89 18 8,48 2,57
test 3 2,19 1,65 0,67 test 3 4,67 3,7
Lab 3, test 1 0,72 0,08 Lab 3, test 1 27 1,55
test 2 0,15 0,87 0,09 test 2 4,46 18,68 1,54
test 3 0,28 0,93 0,54 0,156 test 3 5,4 18,82 6,79 1,54
Lab 4, test 1 1,1 5,7 34,6 0,54 Lab 4, test 1 8 22,7 8,98
test 2 50,8 5,7 30,77 0,495 test 2 17,2 23,5 11,5
test 3 44,2 4,5 40,3 0,605 test 3 6,27 22,2
Lab 5, test 1 3,5 5,3 2,7 1 Lab 5, test 1 0,99 6,33 2,19 2,8
test 2 2,7 5,7 2,9 0,8 test 2 1,56 5,79 2,77 2,44
test 3 5,6 2,2 1 test 3 6,48 2,92 3,39
Lab 6, test 1 3,04 Lab 6, test 1
test 2 4,49 test 2
test 3 3,84 test 3
Lab 7, test 1 0,9 1,5 0,4 0,3 Lab 7, test 1 5,52 22,9 9,69 2,5
test 2 0,9 1,1 0,5 0,3 test 2 7,6 22,9 10,6 3,8
test 3 0,6 1,2 0,5 0,27 test 3 6,15 24,2 9,23 0,13
Lab 8, test 1 0,89 4 0,4 Lab 8, test 1 54 49
test 2 0,92 3,81 0,8 test 2 141
test 3 50 2,62 0,86 test 3 65 84

Table 6: Individual results for method R1 (Resistivity in ohm.cm) and for method prEN
13396 (% Cl) for the increment nº 2 at the duration of the tests of 28 days.
Method R1 (ohm.cm) Method prEN 13396 - 28 d - increment 2 (%)
Laboratory Mix 1 Mix 2 Mix 3 Mix 4 Laboratory Mix 1 Mix 2 Mix 3 Mix 4
Lab 1, test 1 36680 6759 19300 46800 Lab 1, test 1 0,086 0,211 0,167 0,15
test 2 36383 6521 22100 39800 test 2 0,096 0,174 0,137 0,2
test 3 32830 6787 22900 41300 test 3
Lab 2, test 1 25122 5814 22910 34242 Lab 2, test 1 0,165 0,241 0,094 0,29
test 2 26789 5176 17630 38877 test 2 0,215 0,208 0,1145 0,27
test 3 26174 5883 19247 36691 test 3
Lab 3, test 1 18300 5970 155400 28100 Lab 3, test 1 0,109 1,3 0,19 0,7
test 2 15500 9020 158700 29900 test 2 0,058 0,59 0,17 0,9
test 3 23700 9960 158700 8800 test 3
Lab 4, test 1 44851 17849 28713 47015 Lab 4, test 1 0,115 0,354 0,163 0,277
test 2 40604 17058 29074 48694 test 2 0,191 0,328 0,13 0,236
test 3 42597 21237 29629 45245 test 3
Lab 5, test 1 40663 5651 37120 40293 Lab 5, test 1 0,103 0,138 0,117 0,278
test 2 39203 5286 43648 39550 test 2 0,062 0,152 0,099 0,285
test 3 34303 4946 38673 34048 test 3
Lab 6, test 1 Lab 6, test 1 0,14 0,043 0,292
test 2 test 2 0,095 0,083 0,352
test 3 test 3
Lab 7, test 1 34701 7570 25169 31817 Lab 7, test 1 0,096 0,154 0,113 0,097
test 2 30919 7605 26038 30650 test 2 0,081 0,263 0,113 0,079
test 3 25647 7563 22905 36257 test 3
Lab 8, test 1 41850 8700 30680 Lab 8, test 1 0,123 0,237
test 2 42100 8350 28090 test 2 0,141 0,237
test 3 37760 8410 27560 test 3
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 10 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

Table 7: Individual results for method prEN 13396 (% Cl) for the increment nº 2 at the
duration of the tests of 90 and 180 days.
Method prEN 13396 - 90 d - increment 2 (%) Method prEN 13396 - 180 d - increment 2 (%)
Laboratory Mix 1 Mix 2 Mix 3 Mix 4 Laboratory Mix 1 Mix 2 Mix 3 Mix 4
Lab 1, test 1 0,171 0,168 0,194 Lab 1, test 1 0,271 0,202
test 2 0,209 0,171 0,176 test 2 0,238 0,169
test 3 test 3
Lab 2, test 1 0,231 0,259 0,191 0,33 Lab 2, test 1 0,24 0,2811 0,191 0,346
test 2 0,2108 0,2642 0,135 0,34 test 2 0,2918 0,2686 0,1352 0,335
test 3 test 3
Lab 3, test 1 0,148 0,74 0,27 0,8 Lab 3, test 1 0,148
test 2 0,177 0,38 0,24 0,8 test 2 0,177
test 3 test 3
Lab 4, test 1 0,289 0,52 0,324 Lab 4, test 1 0,337 0,62 0,2189
test 2 0,272 0,567 0,2402 test 2 0,337 0,613 0,2334
test 3 test 3
Lab 5, test 1 0,125 0,176 0,163 0,356 Lab 5, test 1 0,176 0,174
test 2 0,163 0,185 0,17 0,321 test 2 0,165 0,243
test 3 test 3
Lab 6, test 1 0,133 0,198 0,384 Lab 6, test 1 0,199 0,21
test 2 0,177 0,22 0,328 test 2 0,16 0,252
test 3 test 3
Lab 7, test 1 0,329 0,218 0,231 0,155 Lab 7, test 1 0,5215 0,237 0,1909
test 2 0,3 0,217 0,218 test 2 0,3801 0,2108 0,1804
test 3 test 3
Lab 8, test 1 0,166 0,31 Lab 8, test 1 0,278 0,231 0,154
test 2 0,192 0,315 test 2 0,25 0,17 0,167
test 3 test 3

Tables 8 to 13 present the cell mean values and the intracell standard deviations, derived from
tables 3 to 7, for the different methods, calculated according to procedures given in the
International Standard ISO 5725-2:1994 [8].

Table 8: Cell mean values and intracell standard deviations for the method D2 (NT BUILD
443) corresponding to the Dns “effective chloride transport coefficient” and K,
“penetration parameter”.

D2-D Cell means D2-D Spread within cells


Laboratory Mix 1 Mix 2 Mix 3 Mix 4 Laboratory Mix 1 Mix 2 Mix 3 Mix 4
L1 2,690 17,167 8,100 2,100 L1 0,915 0,929 2,287 0,361
L2 1,710 17,460 5,363 1,567 L2 0,265 2,781 0,422 0,431
L3 1,357 15,570 4,233 3,650 L3 0,490 2,763 0,660 0,775
L4 1,610 7,377 3,190 L4 0,624 2,436 1,350
L5 1,017 29,377 8,920 2,493 L5 0,287 10,028 5,131 0,669
L6 3,717 45,767 2,973 L6 0,565 18,294 0,305
L7 1,810 14,013 6,367 2,223 L7 0,014 5,157 0,692 0,233
L8 2,307 12,160 11,670 L8 0,159 2,042 3,610
D2-K Cell means D2-K Spread within cells
Laboratory Mix 1 Mix 2 Mix 3 Mix 4 Laboratory Mix 1 Mix 2 Mix 3 Mix 4
L1 24,000 61,333 44,667 22,333 L1 3,606 1,155 6,807 1,528
L2 17,700 57,243 31,863 17,200 L2 1,418 4,633 1,185 2,589
L3 15,543 53,023 28,327 25,557 L3 2,535 4,751 2,085 2,741
L4 16,910 36,703 24,827 L4 3,283 6,369 6,278
L5 15,667 76,667 43,667 23,000 L5 1,528 12,097 10,693 1,000
L6 26,160 237,000 23,587 L6 1,975 47,697 1,237
L7 21,500 54,000 37,333 23,667 L7 0,707 10,000 1,155 1,155
L8 23,000 48,000 52,667 L8 0,000 3,606 9,504
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 11 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

Table 9: Cell mean values and intracell standard deviations for the method M3 (INSA steady-
state-migration).
M3 Cell means M3 Spread within cells
Laboratory Mix 1 Mix 2 Mix 3 Mix 4 Laboratory Mix 1 Mix 2 Mix 3 Mix 4
L1 0,680 0,923 L1 0,113 0,159
L2 1,593 1,668 1,627 1,623 L2 0,042 0,225 0,085 0,067
L3 0,115 0,227 0,177 0,080 L3 0,030 0,025 0,057 0,020
L4 68,437 63,790 49,493 0,431 L4 40,205 0,475 15,698 0,058
L5 0,867 5,233 1,700 0,667 L5 0,115 0,551 0,346 0,252
L6 1,333 2,913 L6 0,217 0,110
L7 0,340 0,580 0,290 0,247 L7 0,036 0,070 0,026 0,045
L8 1,757 2,453 1,407 L8 0,446 0,349 0,492

Table 10: Cell mean values and intracell standard deviations for the method M4 (NT BUILD
492).
M4 Cell means M4 Spread within cells
Laboratory Mix 1 Mix 2 Mix 3 Mix 4 Laboratory Mix 1 Mix 2 Mix 3 Mix 4
L1 2,767 15,433 7,733 2,667 L1 0,208 1,172 0,751 0,462
L2 2,767 16,033 7,500 1,940 L2 0,603 0,839 0,800 0,352
L3 2,467 15,800 4,800 2,900 L3 0,404 1,044 1,114 0,656
L4 1,900 15,167 3,700 2,200 L4 0,100 3,044 0,529 0,265
L5 3,500 16,867 4,133 2,333 L5 0,755 1,041 1,617 0,153
L6 2,163 4,067 29,033 L6 0,670 0,850 45,095
L7 3,033 14,867 5,300 2,667 L7 0,058 0,058 0,458 0,404
L8 2,300 16,733 4,633 L8 0,300 3,972 1,644

Table 11: Cell mean values and intracell standard deviations for the method M6 (Multi-
regime method) corresponding to the Ds (steady state diffusion coefficient) and
Dns (non-steady state diffusion coefficient).
M6-Ds Cell means M6-Ds Spread within cells
Laboratory Mix 1 Mix 2 Mix 3 Mix 4 Laboratory Mix 1 Mix 2 Mix 3 Mix 4
L1 0,387 1,543 0,340 0,200 L1 0,095 0,163 0,044 0,017
L2 1,800 2,100 1,108 0,640 L2 0,436 0,000 0,517 0,044
L3 0,215 0,840 0,540 0,109 L3 0,092 0,108 0,041
L4 32,033 5,300 35,223 0,547 L4 26,992 0,693 4,795 0,055
L5 3,100 5,533 2,600 0,933 L5 0,566 0,208 0,361 0,115
L6 3,790 L6 0,726
L7 0,800 1,267 0,467 0,290 L7 0,173 0,208 0,058 0,017
L8 0,905 3,477 0,687 L8 0,021 0,748 0,250
M6-Dns Cell means M6-Dns Spread within cells
Laboratory Mix 1 Mix 2 Mix 3 Mix 4 Laboratory Mix 1 Mix 2 Mix 3 Mix 4
L1 5,633 16,300 24,367 3,933 L1 1,710 1,153 18,218 0,153
L2 4,587 15,800 8,875 2,850 L2 0,352 3,111 0,559 0,750
L3 4,930 21,500 6,790 1,543 L3 0,665 4,764 0,006
L4 10,490 22,800 10,240 L4 5,875 0,656 1,782
L5 1,275 6,200 2,627 2,877 L5 0,403 0,363 0,386 0,480
L6 L6
L7 6,423 23,333 9,840 2,143 L7 1,067 0,751 0,697 1,861
L8 54,000 85,000 L8 49,153

Table 12: Cell mean values and intracell standard deviations for the method R1 (Resistivity in
ohm.cm)
R1 Cell means R1 Spread within cells
Laboratory Mix 1 Mix 2 Mix 3 Mix 4 Laboratory Mix 1 Mix 2 Mix 3 Mix 4
L1 35298 6689 21433 42633 L1 2142 146 1890 3686
L2 26028 5624 19929 36603 L2 843 390 2705 2319
L3 19167 8317 157600 22267 L3 4168 2086 1905 11697
L4 42684 18715 29139 46985 L4 2125 2220 461 1725
L5 38056 5294 39814 37964 L5 3331 353 3410 3411
L6 L6
L7 30422 7579 24704 32908 L7 4547 23 1617 2958
L8 40570 8487 28777 L8 2437 187 1670
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 12 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

Table 13: Cell mean values and intracell standard deviations for the method prEN 13396 (%
Cl) for the increment nº 2 at the duration of the tests of 28, 90 and 180 days.
prEN-28-2 Cell means prEN-28-2 Spread within cells
Laboratory Mix 1 Mix 2 Mix 3 Mix 4 Laboratory Mix 1 Mix 2 Mix 3 Mix 4
L1 0,091 0,193 0,152 0,175 L1 0,007 0,026 0,021 0,035
L2 0,190 0,225 0,104 0,280 L2 0,035 0,023 0,014 0,014
L3 0,084 0,945 0,180 0,800 L3 0,036 0,502 0,014 0,141
L4 0,153 0,341 0,147 0,257 L4 0,054 0,018 0,023 0,029
L5 0,083 0,145 0,108 0,282 L5 0,029 0,010 0,013 0,005
L6 0,118 0,063 0,322 L6 0,032 0,028 0,042
L7 0,089 0,209 0,113 0,088 L7 0,011 0,077 0,000 0,013
L8 0,132 0,237 L8 0,013 0,000
prEN-90-2 Cell means prEN-90-2 Spread within cells
Laboratory Mix 1 Mix 2 Mix 3 Mix 4 Laboratory Mix 1 Mix 2 Mix 3 Mix 4
L1 0,190 0,170 0,185 L1 0,027 0,002 0,013
L2 0,221 0,262 0,163 0,335 L2 0,014 0,004 0,040 0,007
L3 0,163 0,560 0,255 0,800 L3 0,021 0,255 0,021 0,000
L4 0,281 0,544 0,282 L4 0,012 0,033 0,059
L5 0,144 0,181 0,167 0,339 L5 0,027 0,006 0,005 0,025
L6 0,155 0,209 0,356 L6 0,031 0,016 0,040
L7 0,315 0,218 0,225 0,155 L7 0,021 0,001 0,009 0,000
L8 0,179 0,313 L8 0,018 0,004
prEN-180-2 Cell means prEN-180-2 Spread within cells
Laboratory Mix 1 Mix 2 Mix 3 Mix 4 Laboratory Mix 1 Mix 2 Mix 3 Mix 4
L1 0,255 0,186 L1 0,023 0,023
L2 0,266 0,275 0,163 0,341 L2 0,037 0,009 0,039 0,008
L3 0,163 L3 0,021
L4 0,337 0,617 0,226 L4 0,000 0,005 0,010
L5 0,171 0,209 L5 0,008 0,049
L6 0,180 0,231 L6 0,028 0,030
L7 0,451 0,224 0,186 L7 0,100 0,019 0,007
L8 0,264 0,201 0,161 L8 0,020 0,043 0,009
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 13 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

6 STATISTICAL ANALYSIS OF THE DATA


The statistical treatment of the data has been carried out according to the International
Standard ISO 5725-2:1994 [8] for the determination of the accuracy (trueness and precision)
of measurement methods and results. Part 2: Basic method for the determination of the
repeatability and reproducibility of a standard measurement method. These results were
double-checked by each laboratory to assure the correctness.

According to this standard, the parameters to be calculated are the mean value (m), the
repeatability standard deviation (sr), the reproducibility standard deviation (sR), and
the relationship between m and (sr), (sR).

As a first step in the statistical treatment, it is necessary to critically examine individual


values (tables 3 to 7) in order to find entries that are considered irreconcilable with the other
data. According to the standard, when several unexplained abnormal test results occur at
different levels within the same laboratory, then, that laboratory may be considered to be an
outlier. It may then be reasonable to discard some or all of the data from such an outlying
laboratory. In addition, the standard points out that, obviously erroneous data should be
corrected or discarded.

For each method, in this first analysis, only obvious individual discordant data have been
taken out because if further treated according to this standard, all the individual data from the
lab would have been discarded, even being only one individual data the aberrant one. When
all the individual data belonging to one lab are discordant from the rest, they have been
examined in the systematic application of statistic tests for outliers included in the standard
(step 2).

Therefore, in this step, the following data have been discarded:

• M6-Ds-Lab8-Mix1-test3
• M6-Dns-Lab4-Mix1-test2
• M6-Dns-Lab1-Mix3-test2
• M6-Dns-Lab8-Mix3-test3
• R1-Lab3-Mix4-test3
• M4-Lab6-Mix4-test2
It is necessary to point out that for two labs it was obvious that when applying method M6
there was a mistake in the calculations. After getting in contact and given the permission of
the responsible people of the laboratories, the data were corrected, being the final values
those given in table 5.

The second step consists in the application of statistic tests for stragglers and outliers. The
three methods recommended by the standard have been used:
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 14 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

a: Graphical consistency technique: Mandel´s h and k statistics.

The Mandel´s h statistic, is the between-laboratory consistency statistic and, for each lab, it is
calculated by dividing the cell deviation (cell mean minus the grand mean for the
corresponding mix) by the standard deviation among cell means.

The Mandel´s k statistic, is the within-laboratory consistency statistic and, for each lab, it is
calculated by dividing the intracell standard deviation by the pooled within-cell standard
deviation for each mix.

The h and k values are plotted for each cell in order of laboratory, in groups for each mix
(and separately grouped for the several mixes examined by each laboratory). Examination of
these plots may indicate that specific laboratories exhibit patterns of results that are markedly
different form the others in the study.

The numerical technique has been applied in the following way:

• If the test statistic is less than or equal to its 5% critical value, the item tested is
accepted as correct.
• If the test statistic is greater than its 5% critical value and less or equal to its 1%
critical value, the item tested is called a straggler, and is marked as it, but it is retained
as correct.
• If the test statistic is greater than its 1% critical value, the item tested is called outlier,
it is marked as it and in general, it is discarded.

The details of these tests are properly described in the Standard ISO 5725. An example of
the application of this test to the D2-D method is given in figures 1 and 2, where the
Mandel´s h and statistic are plotted. According to these figures, the laboratory 6 is an
straggler for the Mandel’s h statistic for the mixes 1and 2. Concerning the Mandel’s k
statistic (figure 2), the laboratory 1 is and straggler and laboratories 6, 5 and 4 are outliers for
mixes 2, 3 and 4 respectively.

In these pictures, the depicted critical values have been those corresponding to 8 labs
participating, as it is the less conservative value, in order to retain as much laboratories as
possible. However when making the calculations, the specific number of labs in each level
(for each mix) have been considered. The results obtained from these techniques, with
detailed indication of straggler and outliers are given in table 14.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 15 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

2,500
2,000
1,500
Mandel´s statistic, h

1,000
mix 1
0,500
mix 2
0,000
mix 3
-0,500 1 2 3 4 5 6 7 8
mix 4
-1,000
-1,500
-2,000
-2,500
Laboratory i

Figure 1: Mandel´s h statistic for the method D2 (NT BUILD 443) corresponding to the
diffusion coefficient, D.

2,500

2,000
Mandel´s statistic, k

mix 1
1,500
mix 2
mix 3
1,000
mix 4

0,500

0,000
1 2 3 4 5 6 7 8
Laboratory i

Figure 2: Mandel´s k statistic for the method D2 (NT BUILD 443) corresponding to the
diffusion coefficient, D.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 16 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

b: Cochran´s test

The Standard ISO 5725 assumes that between laboratories only small differences exist in the
repeatability variances. However, this is not always the case, so Cochran´s test has been
recommended to check the validity of this assumption. The details of this test are properly
described in the Standard ISO 5725, but in summary, a parameter, the Cochran´s test
statistic, C, is calculated and compared with its 1% and 5% critical values with the same
criterion than in the case of Mandel´s statistics. In table 14, straggler and outlier values
according to Cochran´s test have also been specified.

c: Grubbs´s test

The aim of this test is the same than Cochran´s one but regarding the reproducibility
variances by testing the significance of the largest observation (sh), of the smallest
observation (sl), of the two largest observations (dh) and the two smallest observations (dl).
The details of these tests are properly described in the Standard ISO 5725, but in summary,
for these four cases the Grubb´s statistics are computed and they are compared with the
critical values for Grubbs´s test. In table 14, straggler and outlier values according to
Grubbs´s test have also been specified.

The third step consists in the calculation of the values of the repeatability standard deviation
(sr) and reproducibility standard deviation (sR). To do so, it is necessary to calculate
previously the variance of repeatability (sr2) and the variance of reproducibility (sL2),
according to the equations detailed in the Standard ISO 5725.

The statistical values determined according to the standard ISO 5725, as well as the details on
the stragglers and outliers according to the different methods applied and then the
laboratories discarded and retained for the study are presented in table 14.

For a better understanding of table 14, a detailed example of the meaning of the symbols and
abbreviations used (for method D2-D and mix 2) is as follows:

• The number of laboratories, n, that finally has been included for the calculations is 6.
• The laboratory 6 (L6) is an straggler according to the Mandel´s h statistic
• The laboratory 5 (L5) is an outlier according to Cochran´s test and according to
Grubb´s test for the double highest observation.
• The laboratory 6 (L6) is an outlier according to Mandel´s k statistic, Cochran´s test
and according to Grubb´s test for the double highest observation.
• The laboratory 6 (L6) has been rejected for further analysis
• The mean value, m, for the 6 remaining laboratories is : 17.624
• The repeatability standard deviation, sr, is 4.959
• The reproducibility standard deviation, sR, is 7.314
• The repeatability coefficient of variation % sr/m is 28.1
• The reproducibility coefficient of variation % sR/m is 41.5
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 17 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

Table 14: Statistical values determined according to the standard ISO 5725.

Notes on stragglers and outliers:


M-h : According to Mandel´s h statistics
M-k : According to Mandel´s k statistics
M-k-h : According to Mandel´s h and k statistics
C: According to Cochran´s test
G-sl: According to Grubb´s test for the single lowest observation
G-sh: According to Grubb´s test for the single highest observation
G-dl: According to Grubb´s test for the double lowest observation

G-dh: According to Grubb´s test for the double highest observation


LX-1: One obvious individual discordant data have been discarded
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 18 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

The dependence of the repeatability standard deviation (sr) and reproducibility standard
deviation (sR) on the mean value, for the different methods is shown in figures 3 to 7. From
these figures it can be deduced that in most of the cases, a linear relationship can be found.
The equations of the linear regressions for all the cases are given in figures 3 to 7.

D2-D D2-K
Sr = 0,2094 m + 0,0515
8 Sr = 0,4159 m + 0,0383 14 2
R = 0,8996
sr 2 sr
7 R = 0,9968 12
Sr Sr
6 10
5
8
sr, Sr

sr, Sr
4
6
3
sr = 0,2848 m + 0,1048 4
2 2 sr = 0,1308 m + 0,1155
R = 0,975 2
1 2 R = 0,8766
0 0
0 5 10 15 20 0 20 40 60 80
m m

Figure 3: Repeatability (sr) and reproducibility (sR) standard deviations in function of the
mean value obtained for the method D2 (NT BUILD 443) corresponding to the
Dns “effective chloride transport coefficient” and K, “penetration parameter”.

Sr = 0,2036 m + 715,38 R1 R1-D


2
R = 0,7736 sr Sr = 0,2045 m + 0,0379
10000 sr 0,7 2
Sr R = 0,9662
Sr 0,6
8000
0,5
6000 0,4
sr, Sr

sr, Sr

sr = 0,0682 m + 423,36
4000 2 0,3
R = 0,9304
0,2
sr = 0,1127 m - 0,0041
2000 2
0,1 R = 0,9778
0 0
0 10000 20000 30000 40000 50000 0 1 2 3 4
m m

Figure 4: Repeatability (sr) and reproducibility (sR) standard deviations in function of the
mean value obtained for the method R1 (Resistivity) and for the corresponding
diffusion coefficients, R1-D.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 19 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

M3 M4
Sr = 0,1009 m + 0,5946
2 sr 2,5 sr 2
Sr = 1,2127 m - 0,3898 R = 0,6537
Sr 2 Sr
R = 0,866
1,6 2

1,2 1,5
sr, Sr

sr, Sr
0,8 sr = 0,1151 m + 0,082 1
2
R = 0,3856 sr = 0,1165 m + 0,2456
0,4 2
0,5 R = 0,964

0 0
0,5 1 1,5 2 0 5 10 15 20
m m

Figure 5: Repeatability (sr) and reproducibility (sR) standard deviations in function of the
mean value obtained for the methods M3 (INSA steady-state-migration) and M4
(NT BUILD 492).

Sr = 0,6482 m + 0,175
M6-De M6-Da
2 sr
2,5 R = 0,9667 8 Sr = 0,3788 m + 0,5347
sr
7 Sr 2
R = 0,919
2 Sr
6
1,5 5
sr, Sr

sr, Sr

4
1 sr = 0,122 m + 0,102
2 3
R = 0,7156
2
0,5 sr = 0,094 m + 0,8202
1 2
R = 0,7593
0 0
0 1 2 3 4 0 5 10 15 20
m m

Figure 6: Repeatability (sr) and reproducibility (sR) standard deviations in function of the
mean value obtained for the method M6 (Multi-regime method) corresponding to
the De (steady state diffusion coefficient) and Da (non-steady state diffusion
coefficient).
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 20 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

pren-28d-pt2 pren-90d-pt2
sr
0,1 0,3
sr Sr = 0,4934 m - 0,0211 Sr Sr = 0,8438 m - 0,1113
2 2
0,08 Sr R = 0,8596 0,25 R = 0,9638

0,2
0,06
sr, Sr

sr, Sr
sr = 0,0448 m + 0,0194
0,15
2
R = 0,0961 Sr = 0,0279 m + 0,0339
0,04
0,1 R2 = 0,0075
0,02 0,05

0 0
0,05 0,1 0,15 0,2 0,25 0,1 0,2 0,3 0,4 0,5
m m

pren-180d-pt2
0,12 sr
Sr = 1,0599 m - 0,175
Sr 2
0,1 R = 0,9072

0,08
sr = -0,0342 m + 0,0344
sr, Sr

0,06 2
R = 0,0062
0,04

0,02

0
0,15 0,2 0,25 0,3
m

Figure 7: Repeatability (sr) and reproducibility (sR) standard deviations in function of the
mean value obtained for the method prEN 13396 (% Cl) for the increment nº 2 at
the duration of the tests of 28, 90 and 180 days.

7 DISCUSSION
In order to make a pre-evaluation of the methods used to identify the best tests methods for
final evaluation in WP5, four key-parameters have to be taken into account: 1) The absolute
value obtained, 2) The accuracy, 3) The relevance of the information that supply and 4) The
appropriateness concerning hours-man and resources.

Within each of these four parameters, several indicators can be identified. For each of the
key-parameters, those are:

1: The absolute value obtained

o 1-1: Deviation of the mean value with respect to the target value (if there is a
target value).
o 1-2: Significant differences with the average (if there is not a target value and
the rest gives very similar values).

2: The accuracy
o 2-1: According to the repeatability coefficient of variation
o 2-2: According to the reproducibility coefficient of variation
3: The relevance of the information that they supply
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 21 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

o 3-1: Concerning the diffusion coefficients Ds and Dns


o 3-2: Concerning the acceleration during the test

4: The appropriateness concerning hours-man and resources

o 4-1: Handling after the test: Need of milling, grinding or cutting, splitting,...
o 4-2: Need of Cl- analysis
o 4-3: Difficulty of handling

Each of these indicators have been classified in three categories, divided according to an
incremental favourable significance, in order to make the ranking of the methods objective.

The key-parameters, indicators and limits for quantification are given in table 15. For each
indicator, the worst category has been assigned 1 point, 2 for the medium one and 3 points for
the better characteristics of the method. Therefore, for the first indicator, there is a total score
of 3 points, as the two indicators exclude mutually, 6 for the second one, 6 for the third one
and 12 for the fourth one. Then, it is necessary to average of the weight of each key
parameter, in order all of them to have the same weight. For example a test method having 6
points for a total of 6 will have “1” for that parameter. As another example, 3 points on 6 on
a parameter will be “0.5” of total score for that parameter.

Table 15: Key-parameters, indicators identified and limits for quantification.


Parameters Indicators Quantification
1 point 2 points 3 points total
1 Absolute value 1-1 Deviation of the mean value with (%bias)>50 10<(%bias)<50 10>(%bias) 3
obtained respect to the target value or impossible
to calculate
1-2 Significant differences with the (%bias)>50 10<(%bias)<50 10>(%bias)
average (if there is not a target value
and the rest gives very similar
values)
2 Accuracy 2-1 According to the repeatability r/m(%)>25 10<r/m(%)<25 10>r/m(%) 6
coefficient of variation
2-2 According to the reproducibility R/m(%)>60 30<R/m(%)<60 30>r/m (%)

3 Relevance of the 3-1 Concerning the diffusion None One of them Both 6
information coefficients Ds and Dns
supplied 3-2 Acceleration Temperature Electrical None
field/Concentrati
on
4 Appropriateness 4-1 Handling after the test: Need of Milling/ Cutting/splitting None 12
milling, grinding or cutting, splitting grinding
4-2 Need of Cl- analysis Solid samples Liquid samples No
Colorimetric
4-3 Difficulty of handling Cell and Cell No
chemical
handling
4-4 Time consumption > 15 days 8 h > t < 15 d < 8 hours
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 22 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

Application to these criteria to the methods

Concerning the first parameter (absolute value obtained) there is not a definite reference
to be used as a target value, as in the case of steady state tests, no natural tests have been
performed. In the case of non-steady-state tests, a natural method have been carried out;
however, it is known that in non stationary tests, different experimental conditions, even in
natural diffusion can lead to different results. So, the “true value” to be considered in each
case is going to be the mean value obtained.

In this way, the mean values of the diffusion coefficients obtained for the three mixes
obtained by the different methods used are given in figure 8 (a-b) for steady-state and non
steady-state coefficients respectively. In the case of M-R, the steady-state coefficient have
been calculated from the equation proposed in the recipe M-R by the IETcc and properly
described in [2,3].

4 20
M3 b) non-steady-state D2-D
a) steady-state M6-De M4
M-R1 M6-Dns
3 15
Dns (E12) (m2/s)
Ds (E12) (m2/s)

2 10

1 5

0 0
Mix 1 Mix 2 Mix 3 Mix 4 Mix 1 Mix 2 Mix 3 Mix 4

Figure 8 (a-b): Mean values of the diffusion coefficients obtained for the three mixes
obtained by the different methods

The percentages of bias with respect to the average for each mix and method are given in
table 16, where, the average (in absolute value) is also given with quantification of the
indicator 1-2 purposes.

From the results in figure 3 and in table 16 it can be noticed that for both transport regimes,
the mean of the coefficients obtained by the three methods are similar enough. All the
methods give values in the same range of magnitude and they discriminate well between the
different types of concretes, being all of them assigned with 2 points as long as indicator 1-2
is concerned 10<(%bias)<50.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 23 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

Table 16: Percentages of bias with respect to the average for each mix and average of bias for
each method.
mix 1 mix2 mix 3 mix 4 average
steady-state average 0,937 2,577 0,942 0,661
bias (%)
M3 6,431 -28,296 10,457 50,213 24
M6-Ds 23,838 12,709 7,171 -31,492 19
R1 -30,269 15,587 -17,628 -18,721 21
non-steady-state average 3,252 17,077 7,006 2,592
bias (%)
D2-D -37,376 3,204 6,091 0,305 12
M4 -19,677 -7,229 -22,924 -3,304 13
M6-Dns 57,053 4,025 16,833 2,999 20

Concerning the prEN method, the mean of the results obtained are given in figure 9, where it
can be seen that, with the exception of mix 1, the values of concentration obtained after 180
days are smaller than those registered after 90 days, which is not the expected behaviour. In
this case, it is not possible to calculate the bias, as there is no other method giving those kind
of results. Therefore, this is assigned with 1 point as long as indicator 1-2 is concerned.

0,5
prEN-28d-2
prEN-90d-2
0,4 prEN-180-2

0,3
% Cl

0,2

0,1

0,0
Mix 1 Mix 2 Mix 3 Mix 4

Figure 9: Mean values of the results obtained using the prEN 13396 standard.

Concerning the second key-parameter (values of accuracy),

The averaged coefficients of reproducibility and repeatability for every method are given in
table 14. For the sake of clarity, the mean values are also depicted in figure 10.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 24 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

100,0

% sr/m
80,0 % sR/m
sr/m, Sr/m ( %)

60,0

40,0

20,0

0,0
D2-D D2-K R1 M3 M4 M6-De M6-Da Pren-28-2 Pren-90-2 Pren-180-2

Figure 10: Coefficients of variations, of reproducibility and repeatability, mean for every
method.

From figure 10 it can be deduced than in general, the accuracy of these methods is not very
good, mainly as long as reproducibility conditions are concerned.

As expected, the values for sr/m (%) are much higher than sR/m (%). Concerning the
coefficient of variation for repeatability values, the highest one is found for the diffusion
coefficient using the method D2, that reaches almost a 30%, being the lowest one for the
measurement of the resistivity of concrete with a value of sr/m smaller than a 10%.

For the coefficient of variation for reproducibility values, the highest values can be found
when calculating the steady state diffusion coefficients in migration cells (M3 and M6-De).
In fact, in table 14 it can be observed that one of the laboratories (L4) has been rejected when
applying the M3 method for three mixes. In addition, it has also been rejected for mixes 1 and
3, and was an straggler for mix 2 when determining the steady-state coefficient by M6. This
has been attributed to the fact that these two methods (M3 and M6) involve a higher degree
of “chemical” handling than the other methods. Therefore, provided that the repeatability
values are quite good, it seems to be necessary a more detailed explanation in the recipes of
these two methods as long as procedures (taken of samples, calibration of conductivimeter,
measurement of effective potential, …) and calculations are concerned.

According to the indicators selected and given in table 15, the score assigned to each method
is as follows:

For the indicator 2-1 based on the repeatability:


• r/m (%) > 25% : D2-D 1 point
• 10< r/m (%) < 25% : D2-K, M3, M4, M6, PrEN 2 points
• 10> r/m (%): R1 3 points
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 25 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

For the indicator 2-1 based on the reproducibility:


• r/m (%) > 60% : M3, M6-Ds 1 point
• 30< r/m (%) < 60% : D2-D, M6-Dns, PrEN 2 points
• 30> r/m (%): D2-K, R1, M4 3 points

Third key-point: Relevance of the information that each method provides

Concerning this point, the first indicator (3-1) refers to the diffusion coefficients Ds and Dns.

On one hand, the prEN 13396 standard, only provides specific values of concentration at
definite points, not giving any parameter able to be used to make predictions. Therefore, it is
given 1 point.

On the other hand there are several methods that provides a single diffusion coefficient. They
are: D2, R1, M3 and M4. D2 and M4 give the value of the non-steady state and R1 and M3
give the steady state diffusion coefficient. Therefore, they are given 2 points.

Lastly, the M6 method provides simultaneously the Ds and Dns coefficients and the ratio
between them addresses the binding ability of the matrix, necessary for using these
parameters in service life predictions. Therefore, this method has 3 points for indicator 3-1.

With respect to the indicator 3-2: Acceleration, prEN accelerates diffusion by temperature,
so, it is given 1 point. The rest of them (except R1 that are given 3 points), accelerates by
electrical field and/or high increase of concentration. So, they have 2 points.

The last point to be taken into account is the convenience concerning hours-man and
resources. At this respect, the assignation of methods according to the indicators is as
follows:

Indicator 4-1: Handling after the test: Need of milling, grinding or cutting, splitting.

The most costly methods are those that need milling of the concrete specimens as they need a
lot of hours-man. The methods that needs this kind of techniques and therefore they have the
score of 1 point are the D2 and the prEN 13396 standard

The M4 method needs immediate splitting of the solid sample into two halves: Therefore, it is
assigned 2 points.

The rest of methods: R1, M3 and M6 are given 3 points.

Indicator 4-2: Need of chloride analysis


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 26 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

After milling the samples in methods D2 and the prEN 13396 standard, it is necessary to
analyse the solid samples, that is quite expensive. Therefore, they have 1 point as long as this
indicator is concerned.

M3 needs analysis of liquid samples and the M4 method applies a colourimetric technique of
analysis. Therefore, they are assigned 2 points.

The rest of methods: R1 and M6 do not have analysis, so, they are given 3 points.

Indicator 4-3: Difficulty of handling

The M3 and M6 method involve mounting and preparing a migration cell and also some
chemical handling. So, they are given 1 point. The method M4 involves montage of the cell
but not chemical handling. So, 2 points for this indicator.

D2, R1 and the prEN 13396 standard are the most easy to handle, therefore, they have 3
points.

Indicator 4-4: Time consumption

The D2 and the prEN 13396 standard are natural diffusion methods, and therefore time
consuming. So, 1 point in this indicator.

The M3, M4 and M6 methods are migration methods that involve from more than 1 day to
about 10-12 days. Therefore, they are given 2 points.

The R1 method is almost instantaneous. Therefore, it is given 3 points.

As a summary, the classification of the different methods according to these four different
points is given in table 16, where the points per indicator for each method are given. In
addition, the sum of the points for indicator have been made, and, as explained previously, in
order to balance the weight of the key-points with less indicators, a score per parameter have
been given (max. 1 for each key-parameter). The sum of them, for the four parameters give
the classification of the methods (last row).
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 27 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

Table 16: Application of the analysis of the methods by key-parameters and indicators
according to the criteria of table 15
Parameters Indicators D2 R1 M3 M4 M6 prEN

1 Absolute value obtained 1-1 -------- -------- -------- -------- -------- --------
1-2 2 2 2 2 2 1
Sub-total 2 2 2 2 2 1
Score/param 0.66 0.66 0.66 0.66 0.66 0.33
2 Accuracy 2-1 1 (D) 3 2 2 2 2
2 (K)
2-2 2 (D) 3 1 3 1 (Ds) 2
3 (K) 2 (Dns)
Sub-total 3 (D) 6 3 5 3 (Ds) 4
5 (K) 4 (Dns)
Score/param 0.5 (D) 1 0.5 0.83 0.5 (Ds) 0.66
0.83 (K) 0.66 (Dns)
3 Relevance of the 3-1 2 2 2 2 3 1
information supplied 3-2 2 3 2 2 2 1
Sub-total 4 5 4 4 5 2
Score/param 0.66 0.83 0.66 0.66 0.83 0.33
4 Convenience 4-1 1 3 3 2 3 1
4-2 1 3 2 2 3 1
4-3 3 3 1 2 1 3
4-4 1 3 2 2 2 1
Sub-total 6 12 8 8 9 6
Score/param 0.5 1 0.66 0.66 0.75 0.5
Total points (max. 4) 2.32 (D) 3.49 2.48 2.81 2.74 (Ds) 1.82
2.65 (K) 2.90 (Dns)

8 CONCLUDING REMARKS AND SUGGESTIONS


According to these results, four methods are revealed as being promising to be retained for
WP5: R1, that has obtained the highest score, followed by M4 and M6 and D2.

Even though it is clear that this classification is not a quantitative one, the obtained results are
quite consistent with the ranking that could have been done by making a qualitative
assessment of the results. In addition, the methods selected are very complementary, as one
of them informs on steady state (R1), two of them on the non-stationary period (M4 and D2)
and M6 makes a conjunction of both.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 28 of 28
WP2 REPORT – Pre-Evaluation of Test Methods

REFERENCES

[1] NT BUILD 443: 94, “Concrete, Hardened: Accelerated Chloride Penetration”;

[2] ANDRADE C., CASTELLOTE M., ALONSO C., GONZALEZ C., Non-steady-state
chloride diffusion coefficients obtained from migration and natural diffusion tests. Part
I: Comparison between several methods of calculation, Materials and structures, vol.
33, jan.-feb. 2000, pp 21-28.

[3] ANDRADE C., ALONSO C., ARTEAGA A., TANNER P., Methodology based on the
electrical resistivity for the calculation of reinforcement service life, Proceedings of the
5th CANMET/ACI International Conference on Durability of Concrete, June 4-9, 2000,
Barcelona, Spain, (Ed. by V.M. Malhotra, ACI, 2000), Supplementary paper, pp 899-
915.

[4] TRUC, O., OLLIVIER, J.P., CARCASSÈS, M, A new way for determining the
chloride diffusion coefficient in concrete from steady state migration test, Cement and
Concrete Research, Volume 30, Issue 2, February 2000, Pages 217-226.

[5] NT BUILD 492: 99, “Concrete, Mortar and Cement Based Repair Materials: Chloride
Migration Coefficient from Non-steady State Migration Experiments”.

[6] CASTELLOTE M., ANDRADE C., ALONSO C., Measurement of the steady and non-
steady state chloride diffusion coefficients in a migration test by means of monitoring
the conductivity in the anolyte chamber. Comparison with natural diffusion tests,
Cement and concrete research, vol. 31, nº 10, 2001, pp 1411-1420.

[7] prEN-13396-1: 2002 “Products and systems for the protection and repair of concrete
structures”. Tests methods. Part I: Measurement of Chloride Ion Ingress by diffusion of
repair mortars and concretes”

[8] International Standard ISO 5725-2:1994 (E). Accuracy (trueness and precision) of
measurement methods and results. Part 2: Basic method for the determination of the
repeatability and reproducibility of a standard measurement method.
CHLORTEST – EU Funded Research Project under 5FP GROWTH Programme

Resistance of concrete to chloride ingress


– From laboratory tests to in-field performance

WP3 REPORT – COLLECTION OF


IN-FIELD DATA

Deliverables D9-10
CONTRACT N°: G6RD-CT-2002-00855
PROJECT N°: GRD1-2002-71808
ACRONYM: CHLORTEST
DURATION: January 2003 – December 2005
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 2 of 29
WP3 REPORT – Collection of In-Field Data

PROJECT
COORDINATOR: SP Swedish National Testing and Research Institute (SP) S

PARTNERS: Institute of Construction Sciences “Eduardo Torroja” (IETcc) E


University of Alicante (UoA) E
Chalmers University of Technology (Chalmers) S
Skanska Norge AS (Selmer) NO
Swedish National Road Administration (SNRA) S
Electricité de France (EDF) F
Netherlands Organisation for Applied Scientific Research (TNO) NL
Hochschule Bremen (HSB) D
Slovenian National Building and Civil Engineering Institute (ZAG) SI
Queens University Belfast (QUB) UK
Laboratório Nacional de Engenharia Civil (LNEC) P
Icelandic Building Research Institute (IBRI) IS
National Institute of Applied Science (INSA) F
Laboratoire Central des Ponts et Chaussées (LCPC) F
Valenciana de Cementos, S.A. CEMEX (VCLC) E
Lund Institute of Technology (LTH) S

ACKNOWLEDGEMENT:
The present document is deliverables of Workpackage 3 – “Collection of In-Field Data”. The
consortium members TNO, LNEC, IBRI, EDF, Chalmers, QUB, SP, IETcc, INSA, HSB, ZAG
and LCPC were involved in the work of this part of the project. The work was led by TNO,
assisted by LNEC, IBRI and EDF.

This document was prepared by Manuela Salta (LNEC) and Rob Polder (TNO)

FURTHER INFORMATION:
Regarding this document: Regarding the project in general
Dr Rob Polder Dr Tang Luping
Netherlands Organisation for Applied Scientific SP Swedish National Testing and Research Institute
Research - TNO Box 857
Schoemakerstraat, 97 S-501 15 BORÅS, Sweden
NL-2600 AA DELFT, The Netherlands Tel. +46-33 165138; Fax: +46-33 134516
Tel. +31-15 276 3222; Fax: +31 15 276 30 18 e-mail: tang.luping@sp.se
e-mail: R.Polder@bouw.tno.nl
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 3 of 29
WP3 REPORT – Collection of In-Field Data

TABLE OF CONTENTS

1 INTRODUCTION 5

2 PARTICIPATING LABORATORIES 5

3 DESCRIPTION AND CHARACTERISATION OF DATA COLLECTED 6

4 SELECTION OF DATA FOR MODELS EVALUATION 17

5 FINAL REMARKS 24

6 REFERENCES 24
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 4 of 29
WP3 REPORT – Collection of In-Field Data
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 5 of 29
WP3 REPORT – Collection of In-Field Data

1 INTRODUCTION
According to the Work-plan of the CHLORTEST project, the objectives to be reached by
WP3 – Collection of In-Field Data are:
• Collection of in-field performance data from reinforced concrete structures in marine
and road environment and other type of structures
• Characterisation of environmental conditions,
• Characterisation of concrete composition

To reach these objectives a previous agreement concerning the interpretation and expression
of data was made and it was decided to prepare data sheets to be filled with data by partners.
Partners agreed to contribute data from national projects.

The work developed in this WP was in interrelation with the WP1 and WP4.

2 PARTICIPATING LABORATORIES
The laboratories that have participated in this WP giving data collected from structures or
concrete specimens from research studies are listed in Table 1. All of the laboratories are in
WP3 according to the consortium agreement, except LCPC, who voluntarily participated in
this WP.

Table 1- Laboratories that participated in data collection


Partner Country
EDF France
IBRI Iceland
IETcc Spain
HSB German
LCPC France
LNEC Portugal
CTH Sweden
QUB United Kingdom
SP Sweden
TNO Netherlands
UoA Spain
ZAG Slovenia
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 6 of 29
WP3 REPORT – Collection of In-Field Data

3 DESCRIPTION AND CHARACTERISATION OF DATA


COLLECTED
After agreement on the type of information necessary, excel data sheets were formulated
upon a proposal by SP and sent by WP3 leader to the participants to be filled. Three types of
data sheets were created: one with concrete composition and properties, another with
environmental exposure characterisation and the third with chloride profiles and data
concerning concrete durability related properties tested in the lab. Annex 1 presents these
excel sheets.

It was decided to collect three types of data taking into account the environmental conditions
and the exposure classes specified in the Standard EN 206. Three task groups were created,
one for marine structures, another for road structures and the third for other type of structures.
The three task groups have collected data covering almost the full range of marine and road
environments in EU.

For marine and road environment, some of the data were collected from existing structures
(bridges, wharves, …) and other from laboratory specimens exposed in natural environmental
conditions, from national research projects carried out by the laboratories involved. Another
group of data were collected from other structures, car parks, sewage pipes transporting
seawater and precast concrete columns for electricity lines. Some other data were also
included concerning results obtained from research studies in simulated standard conditions
in the lab with similar concrete compositions that have been exposed in marine conditions.

In the third project meeting in May 2003 the attendants agreed that chloride profiles would be
considered valid for verification of modelling if they contain the following information:
1- Concrete mixture proportions;
2- Laboratory test methods and results for diffusivity;
3- Exposure conditions;
4- Chloride profiles sampled under the same conditions at preferably three concrete
ages: one or two ages younger than two years (for estimating the age effect) and one
or two at ages older than ten years (for verification of the long term prediction).

In the 4th meeting in October 2004 the problem of chloride profiles validation was reanalysed
and again it was reported that sufficient information should be available about concrete,
environment and at least two profiles for two exposure periods.

More than seven hundred profiles were collected, but only few of them fulfilled all the
conditions to be considered valid according to the criteria fixed on meeting of May 2003.
Only SP's profiles fulfilled all the first requirements. However, having in consideration the
less restricted conditions reported on meeting of October 2004, various profiles from TNO,
QUB, LNEC and IETcc can be also considered valid.

It was decided to consider all the data collected, hoping that some other correlations can be
later made, for example from similar concrete conditions exposed in different regions of
Europe.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 7 of 29
WP3 REPORT – Collection of In-Field Data

Table 2 summarises the number of chloride profiles collected by each exposure class
according to EN 206. As a survey of the data collected, a summary for the different
environments, both natural and in the laboratory, is presented in Tables 3 to 10. In each table
the most relevant information is presented concerning the exposure conditions and exposure
class, the concrete (type and content of cement, type and content of additions, w/b,
compressive strength and the chloride profiles (number of profiles and exposure time). From
these tables it is possible to evaluate the data with respect to the validation criteria.

Table 2 - Number of chloride profiles collected from structures and laboratory concrete
specimens in several exposure classes.
Exposure Class Structures Lab concrete Lab specimens in
by EN 206 specimens in natural simulated conditions in
exposure laboratory
XS1 37 105 -
XS2 15 58 8
XS3 72 238 -
XD3 194 65 -
Total 318 466 8

Some of the laboratories gave additional information concerning the data collected in each
case, including specific information about the structures or test concrete specimens
(geometrical characteristic) and exposure conditions and in a few cases about the corrosion
state. This information was given as an additional small report and in some cases, where the
data are obtained from national research projects, a reference was given to the published
papers or reports. These references are reported in the bibliographic references. In some
cases, only internal reports within the national research projects exist; this is indicated by
"Rep".

In Tables 3 to 5 the most relevant characteristics of the data collected are presented from
marine structures, road structures and other structures. Tables 6 to 9 summarise the most
relevant information concerning the collected data from lab specimens exposed in several
natural environments. In Table 10 data from simulated conditions in the lab are reported.

The collected data contain a large number of profiles, with a reasonably to well documented
background from rather young to mature concrete exposed under different conditions (with
exposure time between 0.5 and 42 years). Performance data for concrete with different
compositions, various types of cement, different additions - fly ash, silica fume and blast
furnace slag, in different exposure classes are available in the data collected. Chloride
profiles have a number of data point variable from 4 to more than 10 and for some cases
duplicate profiles exist for the same exposure time. The chloride analyses were made using
potentiometric titration or volumetric titration (Volhard Method)
In fact, some of the collected data enables to compare the performance of the same concrete
in different marine environmental conditions and also in de-icing salt environment. The effect
of the presence of additions on the concrete performance can also be evaluated for different
exposure conditions.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 8 of 29
WP3 REPORT – Collection of In-Field Data

Table 3 - Data collected from structures in marine environment [a) and b)]

a)
Bridge piers
Type of structure (Haringvielt) Quay Wall(Rotterdam) Bridge Wall breakwater

XS3
Class XS3 XS1 XS3/Tidal Splash XS1 XS3
Zone
North 3 levels, + 9
Sea and 14 m
Air T=
Exposure

water above sea Air T= 1ºC Med.


Environmental Data Water Cl-=18 g/L 2º to
Cl-=18 level Cl-=18 to 6ºC Sea
38ºC
g/L g/L

Cem III
Cem III/B Cem IIA-
Type B Cem III/B 42.5 - -
32.5 M 42.5 R
Cement 32.5
Content, Not
Not def Not def 320 310 349 - -
(kg/m3) def.
Blast
Type - - - - -
furnace
Addition
Content,%
- - - 40 - -
Admixture - - - - - -
Aggregate Type gravel gravel gravel basalt basalt basalt
0.45 0.45 0.45 0.34 - -
Concrete

w/b or w/c
18 18
60 (42 67 66
Fcm, MPa (age) 60 (42 y) 71(35y) 80 (design (design
y) (19y) (30y)
value) value)
ExposureTime (year) 42 42 30 19 35 5 13 14
Number of profiles 6 18 6 6 6 1 1 1
Lab/Additional TNO / Rep [13]
TNO / Rep [12,13] IBRI/Rep UoA
Information

b)
Quay wall
Type of structure Storm surge barrier (Scheldt)
(Nordland)
XS2/
Class XS3/Splash XS3/Splash XS3/Tidal
Exposu

Subm
Environmental Data
Cl-=19 g/L North Sea Cl-=18 g/L
re

Cem III/B
Type Cem III/B 42.5
42.5
Cement
Content
300 325 400 350 350
(Kg/m3)
Type - -
Addition
Content,% - -
Admixture - -
Light
Concrete

Aggregate Type gravel Gravel gravel


weight
w/b or w/c 0.54 0.45
fcm, MPa 75 (8y) 61-75 (18-22y)
ExposureTime (year) 8 18 18-22
Number of profiles 4 3 27 21 2 6
TNO [9,10]
Lab/Additional
/ASTM, TNO Rep [11,12,13] / NT Build 492
Information
Resist.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 9 of 29
WP3 REPORT – Collection of In-Field Data

Table 4 - Data collected from road structures

Road
Road
(beams, Road
(beams and Road
Type of structure slabs and
columns)
columns)

Class XD3 XD3 XD3 XD3


Exposure

T= -1.7º to
16 ºC NaCl
Environmental Data - NaCl
RH=81% 3000g/L
Vertical/ Top
Cem I Cem I 52.5
Type Cem II 42.5 Cem I
42.5
Cement
Content
360 300-360 380 230 410 189 195 325 223
(kg/m3)
Type - - - - - Fly ash
Addition Content
- - - - - 21 20 20 30
(%)
Admixture - - - -
Concrete

Crushed
Aggregate Type - basalt Limestone
limestone
w/b or w/c 0.55 0.45-0.50 0.43 0.84 0.48 0.67 0.67 0.45 0.52
fcm, MPa - - 45 26/30 55 22 23 52 43
ExposureTime,(year) 18 25 to 30 Nd 3 3 3 3 3 3
12 (beams);
Number of profiles 28 1 2 1 1 1 1 1
50 (columns)
CTH / Rep
LCPC /[7,8]
Lab/Additional Information ZAG [2] ☼ IBRI
(Dns Mig. Test)

☼ – Corrosion state, cover, carbonation depth, surface treatment (impregnation); NaCl 12000 kg/km.

Table 5 - Data collected from other type of structures


Pipes Columns
Type of structure Car Park Car Park
(precast) (precast)
Class XD 3 XD3 XS2 XS1/XD1
Expo
sure

T= 7ºC to 18º Cl-=20 g/L Distance to Sea


Environmental Data T= 7º C to 18º C
C t= 9ºC -18ºC > 100 m
Cem I Cem I
Type Cem II/A-M 42.5 R Cem III 32.5
Cement CPA ES CPA
Content(kg/m3) - - 490 400
Type Silica Fume - - -
Addition
Content (%) 7.5 - - -
- - - -
Concrete

Admixture
Aggregate Type basalt basalt - -
w/b or w/c - - 0.55 0.47
fcm (Mpa) - - - -
ExposureTime (year) 27 11 14 27 to 62
10
Number of profiles 11 5 18
(T+F)
IBRI / Rep
Lab/Additional 3 compositions with
IBRI/ Rep EDF/Rep EDF/Rep
Information silano (3y of
exposure)
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 10 of 29
WP3 REPORT – Collection of In-Field Data

Table 6 - Data collected from lab concrete specimens in marine environment-Class XS3 [a) to
g)]

a)
Lab Lab
Lab specimens Lab specimens
Type of specimens specimens specimens
(1x0.5x0.12)m (1x0.5x0.12)m
(0.1x0.1x0.4)m (0.1x0.1x0.4)m

Zone Tidal Tidal Splash Tidal


Exposure

Cl- = 20.3 g/L Cl- = 20.3 g/L


Environmental Data SO4=2.8 g/L SO4=2.8 g/L Cl- = 19 g/L Cl- = 19 g/L
pH=8.1 pH=8.1
Type Cem IV 32.5 R Cem IV 32.5 R Cem I 32.5 R Cem I 32.5R
Cement Content
260 340 300 300
(kg/m3)
Type - - - -
Addition Content
- - - -
(%)
Concrete

Admixture - - - -
Aggregate Type Limestone Limestone Limestone Limestone
w/b or w/c 0.65 0.45 0.50 0.50
Fcm, MPa 32 53 34 34
Exposure Time (years) 0.5 to 2 0.5 to 2 1 to 4 1 to 4
Number of profiles 6 6 4 7
LNEC / Rep LNEC / Rep
LNEC / Rep [6] LNEC / Rep [6]
Lab/Additional Information [5] [5]
Diffusion Cell Diffusion Cell
NT Build 492 NT Build 492

b)
Panels
Type of specimens Blocks
(1.4x1.35x0.15)m

Zone Tidal Tidal


Exposure

Environmental Data North Sea Water Cl- = 20 g/L

Type Cem I Cem III/B LH HS OPC SRPC


Cement Content
350 350 200 and 350 200 and 350
(kg/m3)
Type - - -
Addition
Content (%) - - -
Concrete

Admixture Air entrain; plasticizer - -


Aggregate Type - - -
w/b or w/c 0.48 0.5 0.4
fcm (Mpa) 45 35 - -
2 periods 2 periods
ExposureTime (year) 10 10
( not clear ) (not clear )
Number of profiles 2 2 55 56
Lab/Additional Information HSB/ Rep HSB/Rep IETcc
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 11 of 29
WP3 REPORT – Collection of In-Field Data

c)
Cubes
Type of specimens Not available
(0.1x0.1x0.1)m
Tidal
Zone Tidal / Splash
Exposure

Atlantic Ocean
Cl- = 21 g/L (La Rochelle)
Environmental Data
T=1 to 20ºC

Type OPC OPC Cem I 52.5 PM ES CP2


Cement Content
460 460 230 195 223 325 377 470 410 461 360
(kg/m3)
Type - - - Flyash Flyash Flyash Sílica Sílica - - Sílica
Addition Content
- - 20 30 20 9 11 - - 6
(%)
Concrete

Admixture - - Melamine
Aggregate Type Screened Limestone/calcareous
w/b or w/c 0.4 0.4 0.84 0.77 0.52 0.45 0.30 0.23 0.48 0.32 0.36
fcm (Mpa) 66 30 30 23 52 124 133 0.55 0.80 91
ExposureTime (year) 1 to 7 2 and 5
Number of profiles 38 38 1 1 1 1 1 1 2 6 6
Lab/Additional LCPC/ Rep [7,8]
QUB/Rep
Information

d)
Type of specimens Lab specimens

Zone Splash
Exposure

Environmental
Water Cl-=14g /L
Data

Type CEM I 42.5 N


Cement Content
240 250.5 309.6 369.3 391.7 421.3 421.9 437.7 452.9 455.8 505.1 503.4
(kg/m3)
Type -
Addition Content
-
(%)
Admixture Plasticizer
Concrete

Aggregate Typee Granitic


w/b or w/c 0.35
102
fcm (MPa) 21 26 35 41 42 58 54 58 70 60 96
0.5; 0.5
ExposureTime 1,2 2 and 1 and
0.5 1 1 2 and 0.5 and 0.5 1 1
(year) and 5 5 5
5 10
Number of profiles 1 1 1 3 1 3 4 2 1 1 1 1
Lab/Additional SP / Rep [1]
Information NT Build 492
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 12 of 29
WP3 REPORT – Collection of In-Field Data

e)
Type of specimens Lab specimens

Zone Splash
Exposure

Environmental Water Cl-=14g /L


Data

Cement Type CEM I 42.5 N


Content 233.9 352 376 450 400 479 502 530 502 530 348 384
(kg/m3)
Addition Type Microsilica Flyash Microsilica Microsilica+
+ Flyash Flyash
Content 3 5 10 10 5 5 5 5 20 5 4 + 14 5 + 10
(%)
Concrete

Admixture Plasticizer
Aggregate Type Granitic
w/b or w/c 0.35
fcm(MPa) 21 80 65 117 63 112 125 117 98 95 69 84
ExposureTime (year) 0.5 0.5; 0.5 1 0.5; 1 1 1 and 1 1 0.5; 2 and 5 0.5; 2 and 5
2 and 2; 5 and 5 and
and 5 and 5 5
5 10
Number of Profile 1 3 1 2 6 3 1 2 2 1 3 4
Lab/Additional SP / Rep [1]
Information NT Build 492

Table 7 - Data collected from lab specimens in marine environment- Class XS2 (a and b)

a)
Type of specimens Lab specimens ( 20 x 10) cm

Zone XS2
Exposure

Environmental Data T= 2ºC - 15ºC

Type CEM II 52.5R


Cement Content
-
(kg/m3)
Silica fume,
Type - Sílica fume Blast furnace Sílica fume + Fly ash
Addition blast furnace
Content (%) - 3,7 and 10 7, 30 7 7 + 10,20,40
Concrete

Admixture - - - - -
Aggregate Type basalt
w/b or w/c 0.30 – 0.42
fcm (MPa) - - - - -
Exposure Time (year) 1
Number of profiles 2 2; 6 and 2 2 2 2
Lab/Additional Information IBRI / Rep
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 13 of 29
WP3 REPORT – Collection of In-Field Data

b)
Lab
Type of specimens
specimens
Submerged
Class
Exposure

Twelve sites
Environmental Data Cl-=5 – 23
g /L
Type Cem I 42.5
CementContent
450
(kg/m3)
Type -
Addition Content
-
(%)
Concrete

Aggregate Type -
Admixture -
w/b or w/c 0.4
fcm (MPa) -
Exposure Time (year) 1
Profile number 5
Lab/Additional CTH /Rep
Information [3]

c)
Type of specimens Lab specimens

Zone Submerged
Exposure

Environmental
Water Cl-=14g /L
Data

Type CEM I 42.5 N


Cement Content
240 250.5 309.6 369.3 391.7 421.3 421.9 437.7 452.9 455.8 505.1 503.4
(kg/m3)
Type -
Addition Content
-
(%)
Concrete

Admixture Plasticizer
Aggregate Type Granitic
w/b or w/c 0.35
fcm (MPa) 21 26 35 41 42 58 54 58 70 60 102 96
0.5 0.5, 0.5; 0.5 1;2;5 0.5 1,2,5
1 and 1 and 1,2 1 and 1 and
ExposureTime (year) and 1,2 1 and and and and and
10 10 and 5 10 10
1 and 5 10 10 10 10 10
Number of profiles 2 2 2 4 3 2 4 3 2 2 2 4
Lab/Additional SP / Rep [1]
Information NT Build 492
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 14 of 29
WP3 REPORT – Collection of In-Field Data

d)
Type of specimens Lab specimens

Zone Submmerged
Exposure

Environmental
Cl-=14g /L
Data

Type CEM I 42.5 N


Cement Content
233.9 450 376.3 400 479.3 450 530 502 530 348 384
(kg/m3)
Microsilica+ Microsilica+
Type Microsilica Flyash
Flyash Flyash
Addition
Content
3 10 10 5 5 5 5 20 5 4 + 14 5 + 10
(%)
Concrete

Admixture Plasticizer
Aggregate Type Granitic
w/b or w/c 0.35
fcm (MPa) 21 117 65 63 112 125 117 98 95 69 73
1; 1;
0.5;1;
2 0.5 2; 2; 1
0.5 2 ; 5 0.5;1;2;5 0.5;2;5 0.5; 2;5 0.5;2; 5 and
Exposure Time (year) and and 5 5 and
and 1 and and 10 and 10 and 10 10
5 10 and and 10
10
10 10
Profile number 2 2 2 10 6 7 4 4 2 4 6
Lab/Additional SP / Rep [1]
Information NT Build 492
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 15 of 29
WP3 REPORT – Collection of In-Field Data

Table 8 - Data collected from lab concrete specimens in marine environment- Class XS1 [a)
to c)]

a)
Type of specimens Lab specimens

Zone Atmospheric
Exposure

Environmental
-
Data

Type CEM I 42.5 N


Cement Content
240 250.5 309.6 369.3 391.7 421.3 421.9 437.7 452.9 455.8 505.1 503.4
(kg/m3)
Type -
Addition Content
-
(%)
Admixture Plasticizer
Concrete

Aggregate Time Granitic


w/b or w/c 0.35
102
fcm (MPa) 21 26 35 41 42 58 54 58 70 60 96
0.5;
0.5 2 , 5 0.5 1,5
1 and 1 and 0.5, 2 1; 5 2 and 1 and 1 and
Exposure Time (year) 0.5 and and and and
10 10 and 5 and 10 10 10
10 10 10 10
10
Number of profiles 1 2 2 3 4 4 2 3 5 5 2 3
SP / Rep [1]
Lab/Additional
NT Build 492
Information

b)
Type of specimens Lab specimens

Zone Atmospheric
Exposure

Environmental
-
Data

Type CEM I 42.5 N


Cement Content
233.9 286 376.3 400 479.3 450 450 502 530 348 384
(kg/m3)
Microsilica+ Microsilica+
Type Microsilica Flyash
Flyash Flyash
Addition
Content
3 5 10 5 5 5 10 20 5 4 + 14 5 + 10
(%)
Concrete

Admixture Plasticizer
Aggregate Type Granitic
w/b or w/c 0.35
fcm (MPa) 21 45 65 63 112 125 117 98 95 69 84
0.5, 0.5,
0.5 , 1 , 5 1,5 1,5 1,5 1
2 2, 5 0.5, 2,5 0.5; 2; 5
Exposure Time (year) 0.5 2 and and and and and and
and and and 10 and 10
10 10 10 10 10 10
5 10
Number of profiles 1 3 6 10 9 3 10 3 2 6 6
Lab/Additional SP / Rep [1]
Information NT Build 492
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 16 of 29
WP3 REPORT – Collection of In-Field Data

c)
Lab specimens Lab specimens Lab specimens
Type of specimens
(0.1x0.1x0.4)m (0.1x0.1x0.4)m (1x0.5x0.12)m

Atmospheric Atmospheric Atmospheric


Zone
100 m sea 100 m sea 200 m sea
Exposure

Air 1215 mgCl-/m2d Air 1215 mgCl-/m2d Air 68 mgCl-/m2d


Environmental Data 4-20ºC 4-20ºC 12-40ºC
77% HR 77% HR 78% HR
Type Cem IV 32.5 R Cem IV 32.5 R Cem I 32.5 R
Cement Content
260 260 300
(kg/m3)
Type - - -
Addition
Content (%) - - -
Concrete

Admixture - - -
Aggregate Time Limestone Limestone Limestone
w/b or w/c 0.65 0.45 0.50
fcm (MPa) 32 52.5 34
Exposure Time (year) 0.5 to 2 0.5 to 2 1 to 4
Number of profiles 12 12 5
LNEC/ Rep[5] LNEC/ Rep[5] LNEC/ Rep[6]
Lab/Additional Information NT Built 492 NT Built 492 Diffusion Cell
carb. depth(1 y) carb. depth(1 y) carb. depth

Table 9 - Data collected from lab concrete specimens in de-icing salt environment- class XD3

Type of specimens Lab specimens

Class Vertical or Top Surfaces


Exposure

NaCl
Environmental Data
10000 Kg/Km

Type Cem I 42.5


Cement Content
420 399 260
(kg/m3)
Type - Sílica fume -
Addition
Content (%) - 5 -
Concrete

Admixture - - -
Aggregate type
w/b or w/c 0.40 0.40 0.75
fcm (Mpa) 65 78 30
ExposureTime 0.5 , 1 , 2 to 4
number of profiles 65
Lab/Additional Information CTH/ Rep[2]
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 17 of 29
WP3 REPORT – Collection of In-Field Data

Table 10 - Data collected from lab concrete specimens in simulated accelerated conditions in
the laboratory

Cylinders
Type of specimens Lab specimens
(0.1x0.5m)

Class NaCl
NaCl solution
Exposure

Cl- = 10 x sea water 20g/L 5g/L


Environmental Data 7ºC 7ºC
20ºC 20ºC
Cem II AM
Type Cem III 32.5 Cem I 42.5 R
42.5R
Cement
Content
411 385 411 385 450
(kg/m3)
Type Sílica fume Blast furnace -
Addition
Content (%) 7.5 40 -
Concrete

Admixture Plasticizer -
Aggregate type Basalt -
w/b or w/c 0.40 0.45 0.40 0.45 0.40
fcm (Mpa) 57 53 48 46 -
ExposureTime (year) 2.5 1
Number of profiles 2 2 4
Lab/Additional Information IBRI/ Rep - CTH/ Rep

4 SELECTION OF DATA FOR MODELS EVALUATION


Data collected from tables 3 to 10 (grey cells) were assorted and some of them assembled in
three series by WP4 for model evaluation and benchmarking.

Series 1 – Thirteen chloride profiles, from tables 6 a), c) , d), 7 c) , d) , 8a), b) and 9, were
selected and assembled in this series. All the cases have the same type of cement (CEM I).
Different sub-groups can be organized assembling the same concrete composition and
different exposure conditions. For each case one chloride profile is given and data from short
term tests.

Series 2 - This contains four cases selected from cases in Table 3a). The data selected
correspond to the same type of concrete in two exposure conditions - XS3 and three levels in
XS1. For all 4 cases one chloride profile with exposure time between 6 and 8 years and NT
Build 492 at the same age is given.

Series 3 - This contains five cases with four types of concrete, selected from Table 3b) and
Table 6b). These four cases correspond to similar exposure conditions (XS3) and one case in
XS2. For three cases the chloride profiles obtained after a long exposure time (42 year) is
given and also results from additional tests. For the other two cases only concrete resistivity
data are given.

In Figures 1 to 9 the different chloride profiles selected as Series 1 with similar periods of
exposure are represented. The Figures correspond to the profiles assembled in different sub-
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 18 of 29
WP3 REPORT – Collection of In-Field Data

groups to show the influence of concrete composition and of the exposure conditions on the
concrete performance under chloride load.

Figures 1 and 2 present two profiles selected in this series 1 showing the efect of exposure
conditions with the same concrete composition with and without silica fume, respectively.

Figure 3 shows two profiles from a concrete mix (CEM I) in XS3 exposure conditions (splash
and tidal)

Figure 6 show also two profiles with the same concrete mix (CEM I with silica fume or fly
ash) in the same exposure XS2.

Figures 5, 6 and 9 each represent two profiles with the same concrete mix (CEM I with and
without addition of silica fume) in XS2, XS1 and XS3 conditions, respectively. From this
Figure it is possible to see the increase of performance due to the presence of silica fume in
the three environmental conditions.

Figures 7 and 8 present two chloride profiles with the same concrete composition in each one
with (CEM I with silica fume) and (CEM I), respectivly, in XD3 exposure conditions for a
vertical face and horizontal top face. From these two Figures it is possible to see that the
influence of the concrete surface orientation (vertical or horizontal face) is not relevant for
concrete with silica fume, at least for these short periods of exposure.

Figure 10 presents the four chloride profiles selected in Series 2 and shows the influence of
exposure class and location conditions in XS1 for the same type of concrete (CEM III).

Figure 11 presents the profiles selected in Series 3 and shows the influence of the type of
cement on concrete performance under the same exposure conditions XS3.

2,5

2
C, Cl % binder

1,5

0,5

0
0 10 20 30 40
x, mm

XS2 XS3 XS1

Figure 1 – Three chloride profiles with the same concrete composition (CEM I at the same
content) after the same exposure time in different conditions XS3, XS2 and XS1
(series 1).
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 19 of 29
WP3 REPORT – Collection of In-Field Data

1,8
1,6
1,4
C, Cl % binder

1,2
1
0,8
0,6
0,4
0,2
0
0 10 20 30 40
x, mm

XS1 XS2

Figure 2 – Two chloride profiles for the same concrete (with microsilica) in different
exposure conditions (series 1)

1,6
1,4
1,2
C, Cl % binder

1
0,8
0,6
0,4
0,2
0
0 10 20 30 40
x, mm

XS3/Splash XS3/Tidal

Figure 3 – Two chloride profiles for same concrete mix (CEM I) and the same exposure time
in XS3 splash and tidal locations (series 1).
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 20 of 29
WP3 REPORT – Collection of In-Field Data

2,5

2
C, Cl % binder

1,5

0,5

0
0 10 20 30 40
x, mm

Microsilica Fly ash

Figure 4 - Two chloride profiles for two concrete mixes (CEM I with silica fume and with
flyash) after the same exposure time in XS2 conditions (series 1).

2,5

2
C, Cl % binder

1,5

0,5

0
0 10 20 30 40
x, mm

CEM I_Microsilica CEM I

Figure 5 - Two chloride profiles for two concrete mixes (CEM I with and without addition of
silica fume) after the same exposure time in XS2 conditions (series 1).
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 21 of 29
WP3 REPORT – Collection of In-Field Data

1
0,9
0,8
C, Cl % binder

0,7
0,6
0,5
0,4
0,3
0,2
0,1
0
0 5 10 15 20
x, mm

CEM I_Microsilica CEM I

Figure 6 - Two chloride profiles for concrete mixes (CEM I with and without addition of
silica fume) after the same exposure time in XS1 conditions (series 1).

1,4
1,2
C, Cl % binder

1
0,8
0,6
0,4
0,2
0
0 10 20 30 40
x, mm

Vertical Face Horizontal Top

Figure 7 - Two chloride profiles from the same concrete composition (CEM I with silica
fume), after the same time in XD3 exposure conditions. Comparison between
vertical face and horizontal top face (series 1).
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 22 of 29
WP3 REPORT – Collection of In-Field Data

1
0,9
0,8
C, Cl % binder

0,7
0,6
0,5
0,4
0,3
0,2
0,1
0
0 10 20 30 40
x, mm

Vertical Face Horizontal Top

Figure 8 - Two chloride profiles from the same concrete composition (CEM I without
additions), and the same exposure time in XD3 conditions. Results on vertical face
and horizontal top face (series 1).

1,4
1,2
C, Cl % binder

1
0,8
0,6
0,4
0,2
0
0 10 20 30 40
x, mm

CEM I CEM I_Microsilica

Figure 9 - Two chloride profiles with different concrete compositions (with and without silica
fume) and the same exposure time in XS3 conditions (series 1).
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 23 of 29
WP3 REPORT – Collection of In-Field Data

1,6
C, Cl % binder

1,2

0,8

0,4

0
0 10 20 30 40 50 60
x, mm

XS3/Tidal XS1/9m XS1/14m XS1/14m


.

Figure 10 - Four chloride profiles with the same type of concrete and the same exposure time
in different exposure conditions (XS3 and XS1 at three levels) (series 2).

1,80
1,60
1,40
C, Cl %binder

1,20
1,00
0,80
0,60
0,40
0,20
0,00
0 50 100 150 200
x,mm

OPC_350 SRPC_350 CEM III_300

Figure 11 - Chloride profiles from three concrete compositions, during the same exposure
time in similar XS3 conditions (series 3).
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 24 of 29
WP3 REPORT – Collection of In-Field Data

5 FINAL REMARKS

Although not all the data collected fulfilled all the required criteria for validation in general,
they include information from the performance of different concrete compositions (with the
most important characterization parameters known) exposed in different environmental
conditions representative for the different aggressiveness classes due to chlorides in different
locations in Europe. Most of the cases include information about concrete performance
during early periods of exposure, but some data also exist where information after very long
exposure is available. Few data have information correlating the concrete performance under
chloride load with the corrosion state of the reinforcement.

Nevertheless, some of the data collected represent a very useful tool for model validation and
other for correlation with new similar cases. The authors consider that it is important to
spread this information into the scientific community interested in this field. As it was
previously planned, a CD was made with all the information concerning the data collected.
We recognize the relevance of having this data in a more easy and friendly way, for example
creating an access database, to be used by all the scientific community by internet access.

As a last comment it is also important to stimulate the different members of scientific


community dealing with questions about service life to collect useful information related to
the performance of concrete structures. It is important to collect not only the relevant data
concerning the initiation period but also data about the corrosion state and corrosion
propagation (corrosion rate) in different environmental conditions relevant for the
propagation period.

6 REFERENCES
[1] Tang, L. (2003), "Chloride Ingress in Concrete Exposed to Marine Environment – Field
data up to 10 years’ exposure", SP Report 2003:16, SP Swedish National Testing and
Research Institute, Borås, Sweden
[2] Lindvall, A. (2001), Environmental Actions and Response – Reinforced Concrete
Structures exposed in Road and Marine Environments, Publication P-01:3, Department
of Building Materials, Chalmers University of Technology, Göteborg, 2001, 320 pp.
[3] Lindvall, A. (2003), Chloride ingress data from field exposure, at twelve different
marine exposure locations, and laboratory exposure, Publication P-03:1, Department of
Building Materials, Chalmers University of Technology, Göteborg, 2003, 53 pp.
[4] Lindvall, A. (2002), Chloride ingress in a Swedish road environment – Five years
exposure for three concrete compositions, Publication P-02:4, Department of Building
Materials, Chalmers University of Technology, Göteborg, 2002, 42 pp.
[5] Salta,M.; Melo, A.; Estudo da corrosão do betão armado em exposição marítima e
urbana – Relatório de progresso aos 2 anos, Rel 129/04, Departamento de Materias,
LNEC, Lisboa, 2004
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 25 of 29
WP3 REPORT – Collection of In-Field Data

[6] Costa, A.; Durabilidade de estruturas de betão armado em ambiente marítimo, Tese de
douturamento, Universidade Técnica de Lisboa, IST, Lisboa, 1997
[7] Baroghel-Bouny V., Arnaud S., Henry D., Carcassès M., Quénard D., Ageing of
concretes in natural environments: an experiment for the 21st century. iii - durability
properties of concretes measured on laboratory specimens, Bulletin des Laboratoires
des Ponts et Chaussées, n° 241, nov.-dec. 2002, pp 13-59 (available in French and in
English)
[8] Baroghel-Bouny V., Gawsewitch J., Belin P., Ounoughi K., Arnaud S., Olivier G.,
Bissonnette B. Ageing of concretes in natural environments: an experiment for the 21st
century. IV - Results on cores extracted from field-exposed test specimens of various
sites at the first times of measurement, Bulletin des Laboratoires des Ponts et
Chaussées, n° 249, march-april 2004, pp 49-100 (available in French and in English),
[9] Polder, R.B., Stoop, B.Th.J., Visser, J., 1997, Chloride profiles in cores from a quay wall,
TNO Building and Construction report 97-BT-R0571/01, DuraCrete
[10] Polder, R.B., Walker, R., Page, C.L., 1995, Electrochemical Desalination of Cores from a
Reinforced Concrete Coastal Structure, Magazine of Concrete Research, vol. 47, no. 173,
321-327
[11] Polder, R.B., Rooij, M. de, Vries, J. de, Gulikers, J., 2003, Observed Chloride
Penetration in a marine concrete structure after 20 years in North Sea Environment,
Workshop "Risk based maintenance of Structures", TU Delft, 21 January
[12] Rooij, M.R. de, Polder, R.B., 2003, Accuracy of chloride penetration predictions based
on chloride profile analysis, Advances in Cement and Concrete, Proceedings of a
conference held at Copper Mountain, Colorado, August 10-14, Engineering
Conferences International, 79-88
[13] Polder, R.B., Rooij, M.R. de, 2005, Durability of marine concrete structures – field
investigations and modelling, HERON, Vol. 50 (3), 133-143
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 26 of 29
WP3 REPORT – Collection of In-Field Data

ANNEX
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 27 of 29
WP3 REPORT – Collection of In-Field Data

Sheet 1- CONCRETE CHARACTERISATION


Sample_ID c_Temp_cur
Mix_ID pre_cond_c
Designation_No design_c
Date_prod, yyyy-mm-dd required_c
Cement_type day_c
Cement_class fcm_x
Cement_content N_spec_f
CEM_Blaine f_spec_geom
CEM_Loi f_gross_dens
CEM_SO3 fct_2
CEM_Cl- fct_7
CEM_C3S fct_28
CEM_C2S f_age_cur
CEM_C3A f_RH_cur
CEM_C4AF f_temp_cur
Watercontent pre_cond_f
w_c design_f
w_b required_f
k day_f
Agg_type fct_x
Agg_origine N_spec_Ym
D_max Ym_spec_geom
Agg_content Stat/Dyn
Grading curve Ym_2
air entr_type Ym_7
air entr_content Ym_28
air_nature Ym_age_cur
plast_type1 Ym_RH_cur
plast_nature1 Ym_temp_cur
plast_content1 pre_cond_y
plast_type2 design_y
plast_content2 required_y
plast_nature2 day_y
Entrapped air ym_x
Entrained air Type I add_nature_1
target_air Type I add_cont_1
Gross Density Type I add_cont_1%
flow_value Type I add_nature_2
slump_value Type I add_cont_2
flowdin_value Type I add_cont_2%
Concrete Temp Type II add_nature_1
N_spec_c Type II add_cont_1
c_spec_geom Type II add_cont_1%
c_gross_dens k1
fcm_2 Type II add_nature_2
fcm_7 Type II add_cont_2
fcm_28 Type II add_cont_2%
c_age_cur k2
c_RH_cur Na2Oeq_%Binder
Ini_Cl_%Binder
SHEET 2 – EXPOSURE ENVIRONMENT

Exposure Environment of Concrete


Sample_ID
Mix_ID
Designation_No
Date_Exp, yyyy-mm-dd
Exp class acc to EN 206
Type of exposure

Marine environment Road environment


Exposure zone Type of road
Extension of tidal actions, m Number of lanes
Extension of splash zone, m Wideness of each lane, m
Elevation to seawater, m Speed limit, km/h
Distance to sea, m Traffic density, vehicles/day
Mean Cl concentration, gCl/litre Percentage of heavy vehicles, %
Std dev of Cl concentration Distance to the edge of lane, m
Elevation to seawater, m
Metrological data Type of de-icer
Annual min temp, °C Annual Cl from deicer, kgCl/km
Annual max temp, °C De-icing period, month
Max temp in month No. Max de-icing in month No.
Annual precipitation, mm
Annual mean RH, % Other information
Std dev of RH Attach 1
Exposure to rain Attach 2
Exposure to sun Attach 3
Effect of wind Attach 4
Typical speed of wind, m/s Attach 5
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 29 of 29
WP3 REPORT – Collection of In-Field Data

SHEET 3- CHLORIDE PROFILES


Chloride Ingress in Concrete

Sample_ID
Mix_ID
Designation_No
Date_prod, yyyy-mm-dd
Date_Exp, yyyy-mm-dd
Date_Samp, yyyy-mm-dd

Sample characteristics
Structural part examined
Position of exposure surface
Surface treatment
Surface condition
Samp_size, mm
Cover_thick, mm
Crack_width, mm
Type_steel
Corrosion_status
Carb_depth, mm

Test methods
Profiling method
Cl_analysis method
Binder_analysis method
Lab_test method
Test surface
Free Cl in test, gCl/litre

Laboratory test results Chloride profile


D Cs
x10-12 Cl% C, Cl%
2
Test age, days m /s samp x, mm binder

Other information
Attach 1
Attach 2
Attach 3
Attach 4
Attach 5
CHLORTEST – EU Funded Research Project under 5FP GROWTH Programme

Resistance of concrete to chloride ingress


– From laboratory tests to in-field performance

WP4 REPORT – MODELLING OF


CHLORIDE INGRESS

Deliverables D14-15
CONTRACT N°: G6RD-CT-2002-00855
PROJECT N°: GRD1-2002-71808
ACRONYM: CHLORTEST
DURATION: January 2003 – December 2005
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 2 of 112
WP4 REPORT – Modelling of Chloride Ingress

PROJECT
COORDINATOR: SP Swedish National Testing and Research Institute (SP) S

PARTNERS: Institute of Construction Sciences “Eduardo Torroja” (IETcc) E


University of Alicante (UoA) E
Chalmers University of Technology (Chalmers) S
Skanska Norge AS (Selmer) NO
Swedish National Road Administration (SNRA) S
Electricité de France (EDF) F
Netherlands Organisation for Applied Scientific Research (TNO) NL
Hochschule Bremen (HSB) D
Slovenian National Building and Civil Engineering Institute (ZAG) SI
Queens University Belfast (QUB) UK
Laboratório Nacional de Engenharia Civil (LNEC) P
Icelandic Building Research Institute (IBRI) IS
National Institute of Applied Science (INSA) F
Laboratoire Central des Ponts et Chaussées (LCPC) F
Valenciana de Cementos, S.A. CEMEX (VCLC) E
Lund Institute of Technology (LTH) S

ACKNOWLEDGEMENT:
The present document is deliverables of Workpackage 4 – “Modelling of Chloride Ingress”.
The consortium members LTH, Chalmers, INSA, IETcc, SP and EDF were involved in the
work of this part of the project. The work was led by LTH, assisted by INSA, IETcc, SP and
EDF.

This document was prepared by Lars-Olof Nilsson (LTH)

FURTHER INFORMATION:
Regarding this document: Regarding the project in general
Prof Lars-Olof Nilsson Dr Tang Luping
Lund Institute of Technology SP Swedish National Testing and Research Institute
Division of Building Materials Box 857
S-221 00 LUND, Sweden S-501 15 BORÅS, Sweden
Tel. +46-46 222 7408; Fax: +46-46 222 4427 Tel. +46-33 165138; Fax: +46-33 134516
e-mail: Lars-Olof.Nilsson@byggtek.lth.se e-mail: tang.luping@sp.se
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 3 of 112
WP4 REPORT – Modelling of Chloride Ingress

TABLE OF CONTENTS
Page

1 INTRODUCTION 5
1.1 Background 6
1.2 Objectives of WP4 and the work performed 7
1.3 Assessment criteria 8
1.4 The structure of the report 8

2 PHENOMENOLOGICAL DESCRIPTION OF THE CHLORIDE INGRESS


PROCESS 9
2.1 Concrete, as an ingress media 9
2.2 The concrete pore solution 10
2.3 The exposure conditions 10
2.4 The ingress process 10

3 SOME FUNDAMENTAL PARTS OF INGRESS MODELS 12


3.1 Chloride concentrations 12
3.2 Fick’s 1st law of diffusion 14
3.3 Chloride binding/interaction, binding capacity 14
3.4 Mass balance equations 17
3.5 Fick’s 2nd law of diffusion 18
3.6 The error function, and its compliment 19
3.7 The apparent diffusion coefficient Da and apparent surface chloride content Csa 20
3.8 Time-dependency of diffusion coefficients 20
3.9 The Nernst-Planck equation for ion flux 27

4 TYPE OF CHLORIDE INGRESS MODELS 28


A. MODELS BASED ON FICK’S 2ND LAW. 28
B. MODELS BASED ON FLUX EQUATIONS 29

5 CHLORIDE INGRESS MODELS BASED ON FICK’S 2ND LAW 30


5.1 General 30
5.2 Boundary conditions concepts in empirical models 31
5.3 Models A1a: ERFC-model, constant Da & Csa 33
5.4 Models A1b: ERFC-model, Da(t) & constant Csa 35
5.5 Models A1c: ERFC-model, D(t) & Csa 38
5.6 Models A1d: ERFC-model, Da(t) & Csa(t) 41
5.7 Models A2: Analytical solutions, DF2(t) & Csa(t) 43
5.8 Models A3a and A3b: Numerical, DF2(t) & Csa(t) 46
5.9 Conclusions on models based on Fick’s 2nd law 48

6 CHLORIDE INGRESS MODELS BASED ON FLUX EQUATIONS 49


6.1 General 49
6.3 Boundary conditions concepts in physical models 50
6.3 Models B1a: Models based on Fick’s 1st law of diffusion, no convection 53
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 4 of 112
WP4 REPORT – Modelling of Chloride Ingress

6.4 Models B1b: Models based on Fick’s 1st law of diffusion, with convection 56
6.5 Models B2a, b and c: Physical models based on the Nernst-Planck flux equation 59
6.6 Conclusions, models based on flux equations 63

7 SHORT TERM SENSITIVITY ANALYSIS 64


7.1 Introduction 64
7.2 Methodology for sensitivity analysis 64
7.3 Application of the methodology for ERF 69
7.4 Application of the methodology for LEO 72
7.5 Application of the methodology for MS-Diff 75
7.6 Conclusions 78

8 LONG-TERM SENSITIVITY OF ERFC-MODELS 79


8.1 Mathematical expressions 79
8.2 Sensitivity of various parameters for prediction of chloride concentration 80
8.3 Effect of different parameters on the sensitivity 84
8.4 Discussions 84
8.5 Combined uncertainty of models for prediction of chloride concentration 85
8.6 Conclusions 86

9 BENCHMARKING OF MODELS 87
9.1 Objectives and over-view of work performed 87
9.2 Establishment of criteria for benchmarking 87
9.3 Selection of profiles and documentation prepared for the benchmarking 89
9.4 Selection of models 92
9.5 Responses obtained 92
9.6 Comparison of results 94
9.7 Some examples of results 94
9.8 Analysis of all predictions 99
9.9 Final comments on the benchmarking of models 103

10 CONCLUSIONS 104

REFERENCES 107
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 5 of 112
WP4 REPORT – Modelling of Chloride Ingress

1 INTRODUCTION
Chloride ingress into reinforced concrete is one major process that is limiting the service-life
of the structure. The ingress process, however, is not the only important part of service-life
predictions. For instance, without any accurate chloride threshold levels for initiation of
corrosion, chloride ingress is of little significance. In spite of this, this report only deals with
the ingress part of the process.

The ”state-of-the-art” in modelling chloride ingress into concrete was tested at a Nordic (&
fib TG) seminar in May 2001, Nilsson (2001). A test case was defined with a given concrete,
with given properties in a given environment. Exposure data up to 2 years in that
environment was given. From that data it was easy to determine a suitable surface chloride
content in that particular environment for the submerged zone.

Significant scatter was found in predictions of further ingress, by the ”best” prediction
models available, cf. Figure 1.1, even for the submerged zone. For the splash zone the scatter
was fantastic. These results clearly show that we still have a long way to go before chloride
ingress modelling is accurate enough for design applications.

5 5

H4 H4
4 4 SPL 100 years
Chloride [wt-% of binder]

SUB 100 years


Chloride [wt-% of binder]

3 3

2
2

1
1
given
given 0.6-2 y
0.6-2 y 0
0
0 20 40 60 80 100
0 20 40 60 80 100
Depth [mm]
Depth [mm]

Figure 1.1 Predicted chloride profiles from various prediction models for a given concrete
in a well-defined environment and with profiles up to two years given. Prediction
results for 100 years in the submerged (left) and splash zone, Nilsson (2001)
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 6 of 112
WP4 REPORT – Modelling of Chloride Ingress

An analysis of the predictions shows that the differences may very well be explained by the
background exposure data for each prediction model. That data is more or less individual for
each model. Better data and better exchange of data would probably reduce the scatter
significantly. That exercise remains to be done!

A model for predicting chloride ingress into concrete always aims at predicting the chloride
profile C(x,t) or at least the chloride content at the depth of the reinforcement. The output is
always meant to be compared to a “critical chloride content that is relevant for reinforcement
corrosion, see Figure 1.2. In recent models the prediction and the comparison should include
some measure of the uncertainty in the prediction.

Cenv
C(xc,t) ≥ Ccr ?
Cl-
x
xc

Figure 1.2 The comparison between a predicted chloride content at the depth of the
reinforcement and the chloride threshold level for corrosion

What is a ”model”? A model has some input data that include some information on concrete
and some information on the environment. A model is the process of how to arrive into a
predicted chloride profile or chloride content, from that input data. The output should then fit
or explain data from field performance in various environments such as submerged,
tidal/splash, atmospheric, wicking (tunnels, caissons), de-icing (roads, road bridges, parking
decks, stairs), pools etc.

1.1 Background
This report is the end result of WP4 of EU-project G6RD-CT-2002-00855, ChlorTest. The
over-all objectives of the project include:
• to evaluate different laboratory performance test methods in terms of the theoretical basis,
technical feasibility, measurement precision and applicability in practical construction
design and quality assessment.
• to recommend two reliable methods, one reference method and another rapid method, for
testing the resistance of concrete to chloride ingress.
• to collect in-field performance data of chloride ingress and reinforcement corrosion.
• to verify the laboratory performance tests with the collected in-field performance data
using different models, including scientific, empirical and probabilistic approaches.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 7 of 112
WP4 REPORT – Modelling of Chloride Ingress

• to recommend the practical use of the laboratory performance tests and the interpretation
of the test results, including a proposal for acceptance criteria.
The project has 17 partners from all of Europe, including universities, research institutes,
owners, a material supplier and a contractor. The project period is three years, starting from
February 2003.

1.2 Objectives of WP4 and the work performed


The Work Package 4 of ChlorTest has three objectives:
• Analysis of models,
• Correlation of lab tests to in-field performance,
• Bench marking of models (regarding EU standardization needs as in e.g. EN206,
EC2)

In reality these objectives may be summarized in one sentence: Make sure/exemplify that test
results can be used (in practice)!

The work package has been divided into three tasks, 4.1 - 4.3, with the objectives “Critical
evaluation of models” (4.1), “Sensitivity analysis of models” (4.2) and “Benchmarking of
models” (4.3).

The first chapters in this report complete Task 4.1 of WP4. The task 4.1 has been carried out
by partner LTH with the aid of partner INSA. The objective of this part is limited to:
• A critical evaluation; collect and critically analyse existing models

The task 4.1 has been fulfilled by performing three sub-tasks:


4.1.a Pre-select key issues that a model should take into account (regarding demands from
EN206 & EC-2).
4.1.b Collect models from partners & experts; post list on web-page.
4.1.c Theoretical analysis of models with respect to requirements.

Sub-task 4.1.a has been completed at the 2nd meeting of Chlortest in Bremen by a thorough
discussion of a proposal by the working group. The final result is shown in the next section,
as ”assessment criteria” for models. Sub-tasks 4.1.b and 4.1.c are completed by presenting
these chapters to the partners and a number of invited to a workshop in May 2004 in Lund,
Sweden. At the 3rd meeting of the ChlorTest partners, following the workshop, a limited
number of models was to be selected for further treatment in tasks 4.2 and 4.3.

Task 4.2 Sensitivity analysis (by partners INSA (task leader), SP, EdF) on how each model
appraises
• Environmental differences
• Concrete characteristics
• Corrosion onset

Task 4.2 has been dealt with theoretically, first with a probabilistic approach on the first 10
years ingress for a selection of models and then in a general way on the ERFC-models
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 8 of 112
WP4 REPORT – Modelling of Chloride Ingress

Task 4.3 Bench Marking (by partners IETcc (task leader), SP, LTH, EdF) by studying the
model’s abilities to reproduce reality
• Laboratory test results as input
• In-field performance data as comparison

This was done by selecting some field data and ask model developers to make predictions
with their models on these field cases. The prediction results were then to be analyzed by
using a probabilistic approach.

1.3 Assessment criteria


The basis for the analysis of the chloride ingress models are the assessment criteria. After
proposals that have been presented and discussed at the two first project meetings, a final
decision was taken on the assessment criteria:
• Satisfactory interpretation of real behaviour by the models,
• Practicality of models,
• Ability to relate short-term laboratory test results to long-term performance i.e. test
results as input data,
• Models must be “open”: available, transparent (not black box), input data must be
possible to quantify by the user,
• Ability to take into account different environmental conditions

1.4 The structure of the report


First a qualitative description of the ingress process is given (chapter 2). Then some of the
fundamental parts of various chloride ingress models are explained (chapter 3). The available
models are structured with regard to their physical, chemical and mathematical contents in
chapter 4. Then the different types of models are theoretically, critically analysed in chapters
5 and 6. Separate chapters describe the sensitivity analysis (chapters 7 and 8) and the final
benchmarking of models (chapter 9). In chapter 10 the final conclusions are drawn.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 9 of 112
WP4 REPORT – Modelling of Chloride Ingress

2 PHENOMENOLOGICAL DESCRIPTION OF THE


CHLORIDE INGRESS PROCESS
In this chapter a brief, qualitative description is made of the most essential parts of the
chloride ingress process, and the interaction between these parts. The description is
intentionally made in a very brief way just to point out what kind of “reality” ingress models
are expected to quantify.
The description starts with the ingress media, the concrete!

2.1 Concrete, as an ingress media


Chloride ingress models are generally applied to homogeneous, crack-free concrete with a
one-dimensional ingress at a macro-scale. Some exceptions are available, but the main
difference is pure geometrical. Since a 1D model might be applied to 2D or 3D cases,
possibly with cracks, with pure mathematics and geometry the analysis in this work is limited
to 1D ingress models for a crack-free concrete.

The assumption of homogeneity, however, might be too simple, even if the concrete is very
well mixed and compacted. The wall effect will certainly create a binder content profile close
to a cast concrete surface. This will directly influence the shape of the chloride ingress profile
and the total chloride contents. This is especially significant if the depth of penetration is
small, since the influencing depth of a large wall effect is at least half of the maximum size of
stones. Also at greater depths a certain wall effect is influencing the aggregate content and,
consequently, the binder content.

Vertical separation may give differences in w/c and binder content at different vertical levels.
Concrete will change with time and those changes are somewhat different with depth. The
continuous binder reactions will densify concrete with time and, consequently, change the
pore system with time. Changes with depth will depend on the initial curing and the moisture
conditions of the concrete created by the exposure conditions during its service-life.

Concrete will also interact with the surrounding exposure conditions in other ways that will
change the material with time and in different ways at different depths. Example of
mechanisms that will have such an effect, and influence the chloride ingress process, are
drying and wetting causing shrinkage and swelling, carbonation etc.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 10 of 112
WP4 REPORT – Modelling of Chloride Ingress

2.2 The concrete pore solution


All concretes have a “self-desiccation” due to the contraction of the chemically bound water.
This means that the pore system is not saturated. For low w/c concrete this also means that
the initial pore water has an activity, an RH (relative humidity) much lower than 100 %.
Concrete for chloride environments may have a self-desiccation down to some 85-90 % RH,
corresponding to a degree of capillary saturation of the pores, excluding the air pores, of
some 0.8-0.9.

The pore liquid is not pure water, but a strong solution of ions, mainly sodium, potassium and
hydroxides, besides the chlorides, different for different types of binders and concrete
compositions. The self-desiccation will further increase the concentrations of these species.

2.3 The exposure conditions


The boundary conditions for the chloride ingress process are very different in various
environments. The most simple environment, the submerged zone of marine concrete
structures, the main components of the sea water are chloride and sodium, but a number of
other ions are present in small quantities. Depending on the local conditions, the salinity and
the sea water temperature may have annual variations.

Consequently the exposure solution is quite different from the pore solution in the concrete,
not only concerning the chlorides.

In other environments, the exposure conditions are much more complicated. A “constant”
exposure solution may not be present all the time, but occasionally replaced by pure water
from driving rain, or drying conditions, as in the splash zone of marine structures. In
environments with de-icing salts the exposure solution is very strong in sodium chloride for a
short while, then diluted and sometimes replaced by pure water from rain. Most of the time
the concrete surface is exposed to drying conditions.

2.4 The ingress process


For a saturated concrete constantly exposed to a solution of sodium chloride, the difference in
ionic strength between the pore solution and the exposure solution will initiate a transport of
ions. Chloride ions will move into the concrete pore solution and some of the alkalis and
hydroxides in the pore solution will start to leach out. This is popularly expressed as
“diffusion” of ions, but the various ions will interact with each other in such a way that
charge balance is maintained. Flux of one ion is affected by the fluxes of other ions because
of the electrical field created by all the ions.

In a real concrete, the pore system is not saturated with water. Consequently, there will be a
moisture flow into the pore system. Additionally, depending on the environmental conditions,
that moisture flow can vary with time, in magnitude and direction, causing convection of ions
in the pore system and in and out of the concrete. Below a certain moisture content, the
moisture flow cannot carry any ions, possibly causing an accumulation of ions in certain
areas.

The movement of ions in the pore system is accompanied by a significant interaction with the
matrix, sometimes called “binding”, especially for chloride and hydroxide. Because of this
interaction penetration of chloride is significantly delayed. This interaction depends mainly of
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 11 of 112
WP4 REPORT – Modelling of Chloride Ingress

the type and amount of binder in the concrete, but is affected by for instance the temperature
and the concentration of alkalis.

The chloride interaction/binding seems to be reversible to a large extent. That means that if
the concentration of chloride decreases, e.g. if the concrete surface temporary is exposed to
rain or if the concrete temperature increases, part of the binding is lost. Carbonation will also
cause a loss of most of the binding capacity.

In conclusion, chloride ingress is never a simple “diffusion” process due to concentration


differences only and the “chloride binding capacity” is not a simple, constant property.
Chloride ingress models must include a number of assumptions and simplifications and these
must be stated. The problem is to realize the significance of these simplifications! This is
further discussed in later chapters.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 12 of 112
WP4 REPORT – Modelling of Chloride Ingress

3 SOME FUNDAMENTAL PARTS OF INGRESS


MODELS
In this chapter a number of “fundamentals” for chloride ingress models are defined. Some
fundamental parts have been chosen that are frequently used in various models. The objective
is to state what fundamentals could be agreed upon, without discussing specific models and
have this as a background when the different models are presented, analysed and discussed.

3.1 Chloride concentrations


The content of chloride ions in concrete may be expressed in a number of ways. Nilsson et al
(1996) gave an overview that is summarized here.
The most frequently used expressions are:
cI [kg Cl/m3 of solution] [3.1:1]
3
cII [kg Cl/m of pore solution] [3.1:2]
cIII [kg Cl/m3 of material] [3.1:3]
cIV [kg Cl/m3 of solid material] [3.1:4]
cV [kg Cl/kg of cement or binder] [3.1:5]
cVI [kg Cl/kg of gel] (“gel”=aC+wn) [3.1:6]
cVII [kg Cl/kg of concrete] [3.1:7]
cVIII [kg Cl/kg of sample] [3.1:8]

Consequently, giving the amount of chlorides in only [kg/m3], [kg/kg] or [%] might cause
confusion and large errors in trying to use the data. The mass in kg is sometimes replaced by
moles. It does not change the relations and the relation between kg Cl and moles Cl is
obvious. Concentrations in g/l are identical to concentrations in kg Cl/m3 of solution.

Most of the relations between the different concentrations in [3.1:1] to [3.1:8] are simple. The
concentrations cIII to cVIII are related to each other by the dry density of the material γ, dry
density of the solid material ρ, cement content C, degree of hydration α, the moisture content
w and the molar weight of chloride MCl.

The relation between cII and cIII to cVIII is more difficult, however:

cII = cIII/psol [3.1:9]

where the problem is to define the porosity psol. It includes only that part of the porosity that
contains a liquid which acts as a solvent. One might question whether all the pore water
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 13 of 112
WP4 REPORT – Modelling of Chloride Ingress

really acts as a solvent, i.e. psol = total porosity. Recent measurements by Mangat & Molloy
[1995] give some information. They measured the concentration of free chlorides. When
multiplied by the total amount of pore water, this gave more free chlorides than the total
actually present in the samples! If the measured concentrations are correct, it is obvious that
not all the pore water acts as a solvent. The capillary porosity or the empty porosity at 11 %
RH or 45 % RH might be better alternatives until further research shows the relationship.

Chlorides occur in different forms in concrete. Some of the chlorides are free ions dissolved
in the pore solution. The rest of the chlorides are chemically and physically bound to the
reaction products and their surfaces
C = c + cb [3.1:10]
where c and C are concentrations in [kg/m3], [moles/volume] or [kg/kg]. The free chlorides
are usually given per volume of solute, i.e. [kg/m3 solution]. The bound, and total, amount of
chlorides, however, are given per weight of binder or concrete. Consequently, to distinguish
between free and bound chlorides as in [3.1:10] the porosity psol has to be used.

The question is of course how to define the concept “free” chloride! Several options are
possible:

• Free to move?
• Free to be leached out?
• Free to corrode steel?
• Free to be liberated?

The answer to this question is absolutely vital for physical models, cf. Figure 3.1. Without
consensus about this vital question, it is hardly possible to agree upon physical, or any,
models at all.

Total Cl Total Cl
Bound?
Bound?
c free ( x < 0) =
= c free ( x = 0) ?

Free?
Free?
Depth Depth

Fig. 3.1 Alternative divisions of the total amount of chloride into “free” and “bound” in a
chloride ingress profile
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 14 of 112
WP4 REPORT – Modelling of Chloride Ingress

3.2 Fick’s 1st law of diffusion


Traditionally, Fick’s 1st law has been used to describe the chloride flux as a result of pure
diffusion in liquid water, with the free concentration c of dissolved chloride as the transport
potential.

∂c
qCl = − DF 1 [3.2]
∂x

The “diffusion coefficient” DF1 is then defined by equation [3.2] and by the test set-up used
to measure it.

3.3 Chloride binding/interaction, binding capacity


The interaction between chloride and the matrix of cement-based materials is still not very
well understood. Instead, for most applications a “binding isotherm” is used to give the
relation between the free and bound chloride. Here a number of questions are still disputed,
besides the approach by a binding isotherm at all. Since the binding isotherm is a pure
empirical “property”, it must be measured and the effect of a number of parameters must be
quantified.
The shape of the binding isotherm is one such question, cf. Figure 3.3.1. Does most of the
binding really occur at concentrations close to zero, or is the chloride binding significantly
concentration-dependent?

[Bound or total chloride]


cb or ctot Or?

Or?

[free chloride], cf

Fig. 3.3.1 Alternative shapes of a possible “binding isotherm”.

Another important question is the relation between “free” and “bound” chloride, cf. Figure
3.3.2.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 15 of 112
WP4 REPORT – Modelling of Chloride Ingress

Total
Ctot

Free

Bound

[Free Chloride], cf [g/l]

Fig. 3.3.2 The division of the total amount of chloride into “free” and “bound”

The binding capacity is the capacity of a material to bind chlorides when the ion
concentration changes

∂ cb
binding capacity = [3.1:13]
∂c

This is the slope of a “binding isotherm” with linear scales as in figure 3.3.1. The dimension
of the binding capacity depends on the units chosen for cb and c, respectively.
Obviously the binding capacity depends on the concentration, as seen from figure 3.3.1, but is
sometimes assumed to be a constant. As seen the chloride binding capacity is very high at low
chloride concentrations.
One important, but not obvious, question is whether the concentration of chloride in the pore
water is equal to the concentration in the surrounding sea water or exposure solution. Some
recent ideas on determining “free” chloride by leaching in water cause confusion about “free”
and “water soluble” chloride that really question such a statement.

Chloride binding is, from measurements, a function of several parameters: concentration, pH,
temperature, moisture content, gel content, type of binder, binder content, water-cement ratio,
degree of hydration, time etc. The mechanisms behind all these observations are not fully
understood and much more research is needed and measurements must be much better
performed. See also Larsen (1998).

Recently some ”strange” time effects have been observed on chloride binding in submerged
concrete. The surface chloride content has been found to be time-dependent Cs(t)! A longer
exposure time does not only give deeper chloride ingress but also higher chloride content, cf.
Figure 8.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 16 of 112
WP4 REPORT – Modelling of Chloride Ingress

4.5
5y
4 H4
1y.
3.5 0.6y.

Chloride [wt-% of binder] 2y0-50TL


3
2y50-100

2.5

1.5

0.5

0
0 10 20 30 40 50 60
Depth [m m ]

Fig. 3.3.3 Chloride profiles after several exposure times showing higher chloride
concentration with time. Submerged SRPC concrete with w/b=0.40.

These observations cannot be explained by today’s knowledge. The consequences of


alternative causes and how these findings are utilized in chloride ingress models are quite
contradictory, cf. Figure 3.3.4.

8 8

7 7
C(x,t) [wt-% of binder]
C(x,t) [wt-% of binder]

6 6
Cb(t)
5 5
4 4
Cs(t)
3 3
2 2
Cb=const
1 1
Cs =const
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Depth [mm] Depth [mm]

Fig. 3.3.4 The consequences of alternative causes for the findings in Figure 3.3.3: time-
dependent surface chloride content Cs(t) (left) or time-dependent chloride
binding Cb(t) (right).

A time effect on the ”surface chloride content” Cs in an empirical ingress model, that has no
simple explanation, has a significant consequence: a larger depth of penetration than with a
constant Cs. On the contrary, a time effect on chloride binding will give a smaller predicted
depth of penetration, simply because the larger chloride binding capacity with time will retard
the ingress even more!
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 17 of 112
WP4 REPORT – Modelling of Chloride Ingress

3.4 Mass balance equations


The mass balance equation for chloride can be expressed in different ways, depending on
whether separate terms are used for diffusion and binding. The most simple form gives the
balance of the total amount of chloride per unit volume.

qCl Ctot qCl +dqCl

x dx
Figure 3.4.1. A visual description of the mass balance equation

The flux of chloride qCl is different at different positions in the concrete. The difference in
chloride flux to and from an infinitesimal slice of concrete with a thickness dx will change the
total amount of chloride C in such a slice. The change in total chloride content per unit of
time will be equal to the difference in chloride flux to and from the slice, divided by the
thickness of the slice.
Consequently, the mass balance equation will be

∂C ∂q
= − Cl [3.4.1]
∂t ∂x

To get the dimensions correct, the chloride content here is the content per unit volume of
concrete, not the pore volume, and the flux is per unit area of concrete. The negative sign
simply says that the chloride content will decrease if dqCl is positive. For chloride ingress dqCl
is negative and the chloride content will increase with time.

The change in total chloride content can also be split into a change in free chloride dissolved
in the pore water and a change in bound chloride in such a way that equilibrium will be
maintained between free and bound chloride. Since the chloride flux occurs in the pore water,
the change in chloride content will first occur as change in free chloride. An almost “instant”
change in bound chloride will follow, since the rate of chloride binding is fairly high. The
mass balance equation will then be

∂C ∂c ∂cb ∂q
= + = − Cl
dt dt dt ∂x
or [3.4.2]
∂c ∂q ∂c
= − Cl − b
dt ∂x dt
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 18 of 112
WP4 REPORT – Modelling of Chloride Ingress

3.5 Fick’s 2nd law of diffusion


What usually is called Fick’s 2nd law is in reality a mass balance equation. Originally it was
meant for mass balance in solutions, where no binding exists, but it has been widely applied
to mass transport in porous systems, with and without binding.

3.5.1 Fick’s 2nd law of diffusion for a solution


The law of mass conservation in a small volume of a solution gives the changes with time of
the (free) chloride concentration in that unit volume

∂c ∂ qCl
=− [3.5:1]
∂t ∂x

This is the same equation as Equation [3.4.1], with the free chloride content equal to the total
content, and Equation [3.4.2], with the binding term equal to zero.
With the flow description as Fick´s 1st law for diffusion in a solution, [3.2] and with the
diffusion coefficient DF1 as a constant, one gets

∂c ∂ 2c
= DF 1 2 [3.5.2]
∂t ∂x

This looks like the “Fick’s 2nd law”, but it should be noticed that it is applicable for solutions
only!

3.5.2 Fick’s 2nd law of diffusion for a porous material


For a porous material, with binding, Equation [3.4.2] is applicable. By assuming that the flux
can be described by Fick’s 1st law, Equation [3.2], and that D is a constant, one gets

∂C ∂ ⎛ ∂c ⎞ ∂ 2c
= − ⎜ − DF 1 ⎟ = DF 1 2 [3.5.3]
∂t ∂x ⎝ ∂x ⎠ ∂x

This equation is not identical to “Fick’s 2nd law”, since the concentrations are different in the
two terms. C is expressed per volume of concrete but c is per volume of pore solution. A
factor equal to psol, according to Equation [3.1:9], must be inserted. Additionally, C is the
total content but c is the free content. A factor of

∂C ∂c ∂cb ∂c
= + = 1+ b [3.5.4]
∂c ∂c ∂c ∂c

Then the mass balance equation will be

∂C ∂ ∂c DF 1 ∂ ∂C DF 1 ∂ 2C
= DF 1 = = [3.5.5]
∂t ∂x ∂x ⎛ ∂c ⎞ ∂x ∂x ⎛ ∂c ⎞ ∂2x
p sol ⎜1 + b ⎟ p sol ⎜1 + b ⎟
⎝ ∂c ⎠ ⎝ ∂c ⎠

Comparing this equation to the traditional Fick’s 2nd law


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 19 of 112
WP4 REPORT – Modelling of Chloride Ingress

∂C ∂ 2C
= DF 2 2 [3.5.6]
∂t ∂ x

one realize that the two equations are equal, but that the “diffusion coefficient” DF2 in Fick’s
2nd law is different from the diffusion coefficient DF1 in Fick’s 1st law, for a porous material
with binding.

In fact, the total concentrations C in Fick’s 2nd law could be replaced by the free
concentrations c, but the diffusion coefficient still is the same, DF2

∂c ∂ 2c
= DF 2 2 [3.5.7]
∂t ∂ x

3.5.7 The relation between the D:s in Fick’s laws


From Equations [3.5.5] and [3.5.6] the relationship between the two coefficients in Fick’s
laws are clear

DF 1
DF 2 = [3.5.8]
⎛ ∂c ⎞
p sol ⎜1 + b ⎟
⎝ ∂c ⎠

The magnitude of the relationship depends on the porosity and the binding capacity of the
concrete.

3.6 The error function, and its compliment


The mathematical solution to Fick’s 2nd law, Equation [3.5.6], is well known for a semi-
infinite medium, with a constant diffusion coefficient Da=DF2 and a constant surface chloride
content C(x=0, t)=Csa. If the initial chloride content Ci is negligible the solution is the
compliment to the error-function, Crank (1976)

⎛ x ⎞
C ( x, t ) = C sa ⋅ erfc⎜ ⎟ [3.6]
⎜ 2 D ⋅t ⎟
⎝ a ⎠

where t is the exposure time.

This equation gives a “chloride profile” C(x,t)/Cs=erfc(z)


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 20 of 112
WP4 REPORT – Modelling of Chloride Ingress

1
C(x,t)/C s 0.8

0.6

0.4

0.2

0
0 0.5 1 1.5 2
Z=x/sqrt(4Dt)

Figure 3.6 A chloride profile, chloride concentration with “normalized” depth, from the erfc-
solution to Fick’s 2nd law

3.7 The apparent diffusion coefficient Da and apparent


surface chloride content Csa
Frequently, the erfc-solution to Fick’s 2nd law, Equation [3.6], is fitted to measured chloride
profiles, from structures or specimens. In such a curve-fitting, if the exposure time t is
inserted, the best curve-fitting gives two regression parameters, Da and Csa. The index a
means “achieved” or “apparent”. The definitions are
- Da is the “apparent diffusion coefficient” achieved after an exposure time t, assuming
that the diffusion coefficient DF2 in Fick’s 2nd law was constant during the whole
exposure. Then Da equals DF2.

- Csa is the “apparent surface chloride content” achieved after an exposure time t,
assuming that the boundary condition was constant during the exposure. Then Csa
=C(x=0,t).

3.8 Time-dependency of diffusion coefficients


Nowadays, in most chloride ingress models the diffusion coefficient is treated as time-
dependent. The time-dependency is very much different for different types of diffusion
coefficients.

Empirical models in the beginning of the 1990’s were developed to include the effect that
was obvious in data from different exposure times: a time-dependent Da. Several researchers
started to use expressions for the time-dependency like equation [3.8.1].
α
⎛t ⎞
Da = Daex ⎜ ex ⎟
⎝ t ⎠ [3.8.1]

One example of data for 0.6 to 5 years of exposure is shown in Figure 3.8.1.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 21 of 112
WP4 REPORT – Modelling of Chloride Ingress

1.E-11
Achieved diffusion coefficient, 0.6-5 years
Binder
Concrete
-0.523
Fit Concrete
y = 1E-12x
Fit Binder
Da(t)

1.E-12

-0.4162
y = 1E-12t

1.E-13
0.1 1 10 100
Age, t [years]

Fig. 3.8.1 Apparent “diffusion coefficients” plotted against the length of the exposure time

The time-dependency was found to be different in different exposure zones. Part of it could
be explained by laboratory measurements showing a normal densification of concrete,
somewhat different for different binders. The effect has not yet been fully explained. Can we
use the observations for long-term predictions if we cannot understand it?

One possible explanation is the physical change in the surface layer. This effect and its
consequence for the Da(t) are for instance given by Maage & Helland (1991) and Mohammed
et al (2002). Significant parameter studies are reported by for instance Maage et al (1999).
This research demonstrates that the contribution from ongoing hydration to the aging
exponent is in the range of 0.1 – 0.15 for a number of binders (hydration in fresh water for
periods up to one year) while concrete exposed for seawater experienced an aging exponent
in the range of 0.50 to 0.80 (average 0.60).

Changes of the surface layer as a densification will, however, only change the diffusion
coefficient in the surface layer. It should then be smaller in the outer layer than in the bulk. In
all models, however, the time-dependency of the diffusion coefficient is treated as a change
with time of a diffusion coefficient that is constant with depth. A lower D close to the surface
should change the shape of the chloride profile in such a way that it should be steeper with
time in the surface layer. This is not we observe. An explanation as a “densification” due to
contact with sea water, consequently, must be more or less instantaneous in the whole part of
the concrete that has any chloride ingress. Further research is needed to confirm that.

A time-dependent D causes a lot of confusion. On one hand there are observations showing
the apparent Da to be time-dependent. That D is taken from curve fitting data after a certain
length of exposure to the error-function. That means that the DF2 is regarded as constant
through the whole exposure and the D is a kind of an “average D” during that period, cf.
Figure 3.8.2.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 22 of 112
WP4 REPORT – Modelling of Chloride Ingress

1.E-11
Achieved diffusion coefficient, 0.6-5 years
Dapp (t1 ) = D(0, t1 ) Binder
Concrete
-0.523 Fit Concrete
y = 1E-12x
Fit Binder
Da(t)

1.E-12

-0.4162

Dapp (t 2 ) = D (0, t 2 )
y = 1E-12t

1.E-13 texp
0.1 1 10 100
Age, t [years]

Fig. 3.8.2 The time-dependent apparent D after several lengths of exposure during which
the D is assumed constant!

The time-dependency of the constant, apparent D during an exposure from 0 to t must not be
confused with the diffusion coefficient at a certain age t.

Dapp (0, t ) ≠ Dtest (t )


texp age [3.8.2]

The time-dependency of the apparent diffusion coefficient must be explained before it can be
used with confidence in predictions! Additionally, the age-dependency of a D being a
material property must be determined. We need test methods for that!

An attempt to clarify these things is done in the next section.

3.8.1 Time-dependency of instantaneous diffusion coefficients


A number of measurements show that a chloride diffusion coefficient decreases with time. A
reasonable explanation is of course the continuous cement hydration and binder reactions that
densify the concrete with time. A common equation to describe this time-dependency of the
diffusion coefficient D(t) is

n
⎛t ⎞
D(t ) = Dtest ⎜ test ⎟ [3.8.1a]
⎝ t ⎠

where t is the age, ttest is the age at testing, Dtest is the test result at that age.
An example is shown in Figure 3.8.3, with Dtest=10-11 m2/s at ttest=30 days and n=0.50.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 23 of 112
WP4 REPORT – Modelling of Chloride Ingress

1.E-10

D(t)
D(t) [m /s] 1.E-11
2

1.E-12

1.E-13
0.00 0.01 0.10 1.00 10.00 100.00
Age [years]

Fig. 3.8.3 An example of the time-dependency of the chloride diffusion coefficient. The
parameter t is the age of the concrete

Such a time-dependency of the diffusion coefficient means that the diffusion coefficient
continuously changes with time or actually with age. It should be noticed, however, that the
decrease with time is shown for a limited age only, sometimes only for six months and a few
years for certain binders.

3.8.2 Time-dependency of apparent diffusion coefficients


A lot of field and laboratory data also show a clear time-dependency of the apparent diffusion
coefficient Da(t-texp) That time-dependency is usually expressed in terms of the exposure time
t-texp, where texp is the concrete age at exposure.
m
⎛ t ⎞
Da (t exp , t ) = D0 ⎜ 0 ⎟ [3.8.2a]
⎜t −t ⎟
⎝ exp ⎠
where t is the age, texp is the age at exposure, D0 is the diffusion coefficient at a reference age
t0. The exponent m is different from n in equation [3.8.1]. One obvious peculiarity with this
expression is that Da(t-texp) is not defined at the time of exposure!
It is obvious that the Da(t-texp) is, and must be, different from D(t), but the relation between
them is not obvious and among other things dependent on the age at exposure. Additionally,
the explanation of the time-dependency of the apparent diffusion coefficient is not only the
continuous densification of the concrete.
The apparent diffusion coefficient is time-dependent, but it is assumed to be constant during
the complete exposure time, as a consequence of its definition! An example is shown in
Figure 3.8.4a, where the time-dependency of the instantaneous diffusion coefficient in Figure
3.8.3 is used to determine a numerical solution to Fick’s 2nd law.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 24 of 112
WP4 REPORT – Modelling of Chloride Ingress

1.E-10

D(t)
Da(texp, t)
2
D(t) [m /s] 1.E-11

1.E-12

1.E-13
0.00 0.01 0.10 1.00 10.00 100.00
Age [years]

Fig. 3.8.4a An example of the time-dependency of the apparent chloride diffusion


coefficient. The parameter t is the age of the concrete and texp is the age at
exposure, 30 days or 0.082 years.

From Figure 3.8.4a it is obvious that the time-dependency of the apparent diffusion
coefficient does not exactly follow an equation like [3.8.2a]. However, such an equation
gives an approximate description, with a different exponent, of course. An m-value of 0.39
gives a fairly good fit for the example in Figure 3.8.4a, cf. Figure 3.8.4b.
1.E-10
D(t)
Da(texp, t)
1.E-11 Da(t)
D(t) [m 2/s]

1.E-12

1.E-13
0.00 0.01 0.10 1.00 10.00 100.00
Age [years]

Fig. 3.8.4b An example of an approximate time-dependency of the apparent chloride


diffusion coefficient according to an equation like [3.8.2a], for m=0.39.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 25 of 112
WP4 REPORT – Modelling of Chloride Ingress

1.E-11
D(t)
8.E-12 Da(texp, t)
Da(t)
D(t) [m /s]
6.E-12
2

4.E-12

2.E-12

0.E+00
0.01 0.10 1.00 10.00 100.00
Age [years]

Fig. 3.8.4c An example of an approximate time-dependency of the apparent chloride


diffusion coefficient according to an equation like [3.8.2a], for m=0.39. Figure
3.8.4.b, but with a linear scale on the vertical axis.

Some researchers quantified theoretically the relationship between the apparent diffusion
coefficient and the instantaneous one. Visser et al [2002] for example pointed out the
theoretical relationship if they are described by Equations [3.8.1] and [3.8.2a] respectively.

n
D ⎛ t ⎞
Da (t exp , t ) = 0 ⎜ o ⎟ [3.8.2b]
1 − n ⎜⎝ t − t exp ⎟

This equation, however, is not quite realistic. It is based on another assumption, that the
diffusion coefficient is not defined at the time of exposure. It is said to be valid for large
values of t, only.

A correct description of Da(t-texp) can be derived from Crank (1976). For a time-dependent
D(t), the Da(t) times the exposure time t-texp in the erfc-solution is found by integrating D(t)
over time, from the start of exposure

t
Da ⋅ (t − t exp ) = ∫ D(t ' ) ⋅ dt '
t = t exp
[3.8.2c]

With Equation [3.8.1] inserted, the resulting apparent diffusion coefficient will be

t n
1 ⎛ t test ⎞
Da (t exp , t ) =
t − t exp ∫t Dtest ⎜⎝ t ⎟⎠ dt [3.8.2d]
exp

and the expression will be


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 26 of 112
WP4 REPORT – Modelling of Chloride Ingress

1− n
Dtest ⎛ t test ⎞ t − t ⋅ t exp
n n

Da (t exp , t ) = ⎜ ⎟ ⋅
1− n ⎝ t ⎠ t − t exp
or [3.8.2e]

D ⎛t ⎞ ⎛ ⎛ t ⎞
n
n

Da (t exp , t ) = test ⎜ test ⎟ ⋅ ⎜⎜ t − ⎜ ⎟ ⋅ t exp ⎟ ⋅ 1
1− n ⎝ t ⎠ ⎜t ⎟ ⎟ t −t
⎝ ⎝ exp ⎠ ⎠ exp

A similar expression was shown by Gulikers (2004).This expression is shown in Figure


3.8.4d to coincide with the result from curve-fitting the numerical solution in Figure 3.8.4a.
1.E-10
D(t)
Da(texp, t) num
Da(t-texp) eq
1.E-11
D(t) [m /s]
2

1.E-12

1.E-13

0.00 0.01 0.10 1.00 10.00 100.00


Age [years]

Fig. 3.8.4d An example of a time-dependent apparent chloride diffusion coefficient


according to equation [3.8.2e], compared to the numerical solution and curve-
fitting

3.8.5 Conclusions on time-dependent diffusion coefficients


The time-dependency of diffusion coefficients is certainly a possible source of
misunderstanding and even mathematical errors. In any model where such a time-dependency
is used, the time-dependency must be very well defined. Here only the continuous
densification of concrete was taken as a cause of the time-dependency of diffusion
coefficients. Other reasons are possible, and likely, especially under field conditions. Those
reasons, however, are not yet fully explained.

Since the aging factor gets a contribution both from continued hydration and the beneficial,
but not understood, contact with seawater (exponents β and γ, respectively). Thus the correct
argument should be the concrete’s total age for the hydration effect and the time of exposure
for the extra effect of being in contact with seawater. For many practical applications
(assessment of existing structures), these two periods will be more or less the same. However
for parameter studies of the effect of curing conditions etc., we have to split up these two
effects. A possible way of getting a consistent model is therefore to describe the two
components of the aging factor and let them be functions of total age and period of exposure
respectively.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 27 of 112
WP4 REPORT – Modelling of Chloride Ingress

3.9 The Nernst-Planck equation for ion flux


Nowadays it is widely recognized that Fick’s 1st law is a very strong over-simplification of
chloride transport. Instead, the effect of the electric potential field Φ, from other ions and
from an applied potential difference in migration tests, is included in the flux equation. The
flux is then described by the Nernst-Planck equation.

⎛ ∂c ∂ ln ai z i F ∂Φ ⎞
qCl = − Di ⎜ i + ci + ci ⎟ [3.9]
⎝ ∂x ∂x RT ∂x ⎠

The “diffusion coefficient” in Fick’s 1st law DF1 is obviously not a material property but
depends on the conditions. Consequently, it cannot be determined by a simple “diffusion”
test. Any attempt with straight forward diffusion cells will determine something else since
equation [3:2] is not a correct description of what happens in such a test.

Instead it must be acknowledged that the flux of chloride is influenced by the other ions. In
any test set-up for determining the “diffusion coefficient” Di for chloride, using equation
[3:8], the result will depend on the diffusion coefficients for all other ions. Consequently, it is
not possible to determine the chloride diffusion coefficient directly in one simple test.
Different approaches to solve this problem have been used. Truc (2000) made estimations
from chloride migration tests where the diffusion coefficients for sodium, hydroxyl and
potassium ions were adjusted to fit predictions from a multi-species model. Samson (1999)
instead utilized the diffusion coefficients for several ions from data for diffusion in a solution.
By determining the “formation factor”, i.e. the effect of the tortuosity and the restrictivity of
the pore system, he could describe the flux of all ions by applying the same formation factor
to the flux of all ions.

The main difficulty in applying Equation [3.9] is to quantify the potential gradient ∂Φ/∂x
since it is a function of the fluxes of all ions and consequently changes with time. In a
migration test simple versions of Equation [3.9] are used since the applied electrical field
may dominate the potential field.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 28 of 112
WP4 REPORT – Modelling of Chloride Ingress

4 TYPE OF CHLORIDE INGRESS MODELS


Different prediction models have a number of similarities and differences. They require
different input data that look alike, but sometimes are quite different. A structure of the
available methods is shown in this chapter. The various models are described in more details
in the next chapter.

A. MODELS BASED ON FICK’S 2ND LAW.


“Empirical methods” are empirical or semi-empirical methods where chloride profiles are
predicted from analytical or numerical solutions (true or false!) to Fick’s 2nd law of diffusion
or other models where the diffusion coefficient DF2 in Fick’s 2nd law is used, sometimes as an
“apparent diffusion coefficient” Da.

ERFC-methods
A1a Constant Da & Csa (Traditional, Collepardi, Japan, Tuutti)
A1b Da(t) & constant Csa (Selmer-Poulsen, Bamforth, Allied, Firth)
A1c D(t) & constant Csa (DuraCrete)
A1d Da(t) & Csa(t) (Nilsson)

Analytical solutions to Fick’s 2nd law (other than erfc)


A2 DF2(t) & Cs(t) (Mejlbro-Poulsen, Hetek)

Numerical solutions to Fick’s 2nd law


A3a DF2(t) & Csf(t), binding isotherm (NIST, LEO)
A3b DF2(t) & Cs(t) (Life-365)
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 29 of 112
WP4 REPORT – Modelling of Chloride Ingress

B. MODELS BASED ON FLUX EQUATIONS


“Physical methods” are methods where chloride transport and chloride binding are described
with separate, physical expressions.

Models based on Fick’s 1st law


B1a Binding isotherm (ClinConc)
B1b Convection models (HetekConv, Imperial, Toronto, THI EdF, LERM,
Meijers)

Models based on the Nernst-Planck flow equation


B2a Binding isotherm (MS-Diff, Li&Page)
B2b Ion-solid equilibrium (Johannesson)
B2c Chemically bound Cl (Stadium)
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 30 of 112
WP4 REPORT – Modelling of Chloride Ingress

5 CHLORIDEND
INGRESS MODELS BASED ON
FICK’S 2 LAW
5.1 General
Empirical prediction models for chloride ingress into concrete utilize experience to fit data to
mathematical models. If the fit is not good enough, more regression parameters are added.
New experience is required for a new concrete or a new environment.

Predictions by empirical models should be done during the observation period only.
Verification is then achieved from a comparison between data and model, cf. Fig. 5.1.1.
Scatter between observations and predictions give the model uncertainty.
40
Depth [mm] of Penetration
35

30

25

20

15

10

0
0 1 2 3 4 5 6
Exposure Time [years]

Fig. 5.1.1 Predicted chloride ingress by an empirical model in the time interval where the
model is valid: the observation period.

Predictions beyond exposure times where the model has been calibrated the data are highly
questionable, since an empirical model by definition is only a mathematical fit to
observations. Since it really has no clear physical meaning, it is dangerous to use it for
extrapolation outside the borders of the observation data. Figure 5.1.2 shows this danger very
clear. It is an alternative way of showing the predictions in the left part of Figure 1.1.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 31 of 112
WP4 REPORT – Modelling of Chloride Ingress

120

Penetration depth (for C=1.0 % of binder) [mm


100
Alternative
80
predictions:
60
Data
0.6-2 y
40

20

0
0.1 1 10 100
Exposure time [years]

Fig. 5.1.2 The alternative predictions for 30-100 years in Figure 1.1 compared to the
observation data between 0.6 and 2 years. The circle at 5 years represents
later measurements that were not available for the predictions.

5.2 Boundary conditions concepts in empirical models


In empirical models the boundary condition is expressed as a surface chloride content Cs, that
really is the response by a particular concrete to particular environmental actions. That
response does not only depend on the environment but also on the concrete! Consequently, a
Cs cannot be taken as boundary conditions for a particular environment. The effect of the
concrete must be considered as well.

The Cs is the total amount of chloride in the surface-near region of the concrete.
Consequently, it depends on the porosity of the concrete and on the binding capacity of the
concrete. That explains why Cs is a function of the concrete mix, especially the type of
binder, the binder content, and the water-binder ratio. It also explains why Cs is a function of
temperature, pH, carbonation etc.
1.2

1
13 concretes submerged for 5 years
Cl % per sample

0.8
Cs = 0.5 to 1.2( wt-% of concrete)
0.6

0.4

0.2

0
0 10 20 30 40 50 60
Depth(mm)

Fig. 5.2.1 Several examples of Csa = f( concrete mix) in one marine, submerged
environment
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 32 of 112
WP4 REPORT – Modelling of Chloride Ingress

Since the apparent diffusion coefficient obviously depends on the binding capacity, cf.
Equation [3.5.8], there is correlation between the Da and Csa for a particular concrete. They
cannot be chosen independently!

In the 1990’s the Csa has been regarded as time-dependent by a number of researchers. The
explanation for this is not yet given, but in some environments it is obvious that the chloride
content could increase with time, e.g. the splash zone in road environments. Drying and
wetting in the splash zone of marine structures could possible explain a time-dependent Csa in
that zone.

However, recently it was observed that Csa(t) also is time-dependent in the submerged zone
of marine structures, cf. Figure 3.3.3. The effect was not visible if the data is expressed as
chloride by weight of sample simply because of the large scatter, cf. Figure 5.2.2. However,
the time-dependency is clearly visible if Csa is expressed by weight of binder, see Figure
5.2.3. We cannot explain why, however! This is vital, both for empirical models and for
physical models. We urgently need better, long-term data.

1.0

0.8

0.6
Csa(t)

y = 0.0222Ln(x) + 0.6542
0.4

0.2

Achieved Surface Chloride Content


0.0
0.1 1 10 100
Age, t [years]

Figure 5.2.2 Surface chloride contents Csa as a function of exposure times 0.6 to 5 years for
the concrete in Figure 1.1. Csa expressed by weight of sample.

7
Achieved Surface Chloride Content
6
Csa(t) [wt-% of binder]

2 y = 0.5676Ln(x) + 3.5998

0
0.1 1 10 100
Exposure time, texp [years]

Figure 5.2.3 Surface chloride contents Csa as a function of exposure times 0.6 to 5 years for
the concrete in Figure 1.1. Csa expressed by weight of binder.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 33 of 112
WP4 REPORT – Modelling of Chloride Ingress

5.3 Models A1a: ERFC-model, constant Da & Csa


(Traditional, Collepardi model)

The most frequently used empirical model for chloride ingress is the error-function solution
to Fick’s 2nd law. This model was almost the only one for twenty years, between 1970, when
Collepardi [1970] proposed it, up to around 1990. It is still widely used.

5.3.1 Principle
The principle of this model is:

1) Estimate the “apparent diffusion coefficient” Da and the “apparent surface chloride
content” Csa, from field exposure or laboratory tests.
2) Predict the chloride ingress, the chloride profile, from the erfc-solution to Fick’s 2nd law,
with constant parameters.

5.3.2 Mathematics
In full the prediction equation is
⎛ x ⎞
C ( x, t ) = Ci + (C sa − Ci ) ⋅ erfc⎜ ⎟
[5.3.2a]
⎜ 2 D ⋅ (t − t ) ⎟
⎝ a ex ⎠

but it can be shortened, if the initial chloride content Ci is negligible, to


⎛ x ⎞
C ( x, t ) = C sa ⋅ erfc⎜ ⎟ [5.3.2b]
⎜ 2 D ⋅ (t − t ) ⎟
⎝ a ex ⎠

An alternative empirical model that has been proposed is, i.e. by Andrade (2004) :
xc =Ccr = k ⋅ Da ⋅ t [5.3.2c]
It looks much simpler, but it is not! The parameter k can easily be found from equation
[5.3.2a], where it is obvious that the two models really are identical!

The chloride content C is the total chloride content of the particular concrete. However,
alternatively the free chloride concentration in the pore system could be used in the same set
of equation, of course with the Csa replaced by csa of the surrounding environment, Tuutti
[1983].

5.3.3 Required Input data


The model requires only two parameters to be quantified, Da and Csa.

5.3.4 Input relation(s) to test methods


Da can be derived from a short term immersion test, in the laboratory or under field
conditions.

5.3.5 Required boundary conditions


The required boundary conditions are simple, the future achieved response by the particular
concrete in the particular environment.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 34 of 112
WP4 REPORT – Modelling of Chloride Ingress

5.3.6 Model(s) for environmental actions


Environmental actions are translated into a surface chloride content Csa mainly from exposure
tests in the true environment, for the concrete that is going to be used.

Some models for Csa are available, i.e. Lindvall [2003], where Csa can be estimated from the
concrete composition (type and amount of binder, w/c etc.), and various environmental
parameters, such as sea water temperature and salinity, distance to sea level, distance to road
surfaces etc.

5.3.7 Output data


The prediction gives a chloride profile like the erfc-solution for each exposure time.

5.3.8 Openness/Practicality
The model is very simple and fully transparent for any user.

5.3.9 Interpretation of real behaviour


Today it is obvious from numerous field observations that the model tremendously
overestimates the chloride ingress rate. Expressed in terms of achieved diffusion coefficient
Da it is quite clear that Da could not be treated as a constant in time. The model is today
almost always replaced by one of the models where Da is time-dependent.

5.3.10 Advantages
The main advantage is the simplicity in performing the predictions.

5.3.11 Limitations & Drawbacks


Besides the general limitations and drawback for empirical models, see section 5.9, the main
limitation and drawback is the great lack of coincidence between predicted and measured
profiles, also after a few years of exposure, if more than one exposure time is used in the
observations.

5.3.12 References
Collepardi [1970], Tuutti [1983], Poulsen [1990], Andrade [2004]
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 35 of 112
WP4 REPORT – Modelling of Chloride Ingress

5.4 Models A1b: ERFC-model, Da(t) & constant Csa


(The Selmer-Poulsen model etc.)

In the late 1980’s a number of researchers started to apply time-dependent apparent diffusion
coefficients to chloride ingress models. The most elaborated model is the Selmer-Poulsen
model, based on the research of a number of other researchers.

5.4.1 Principle
The principle of this model is:

1) Estimate the time-dependency of the “apparent diffusion coefficient” Da(t) and the
(constant) “apparent surface chloride content” Csa, from field exposure or laboratory tests.

2) Predict the chloride ingress after a certain exposure time, the chloride profile, from the
erfc-solution to Fick’s 2nd law, assuming constant parameters DF2 and Csa during exposure.

5.4.2 Mathematics
The mathematical model simply is the same as the previous one

⎛ x ⎞
C ( x, t ) = C i + (C sa − C i ) ⋅ erfc⎜ ⎟ [5.3.2a]
⎜ 2 D (t ) ⋅ (t − t ) ⎟
⎝ a ex ⎠

where t is the age of the concrete and tex is the age at exposure, tex being negligible at long
exposure times. The time-dependency of the apparent diffusion coefficient Da(t) is found
from empirical equations like

α
⎛t ⎞
Da = Daex ⎜ ex ⎟ [3.8.1] & [5.3.2b]
⎝ t ⎠

where t is the age of the concrete and tex is the age at exposure. Daex is the (undefined)
apparent diffusion coefficient at the time of exposure. The exponent α depends on the
environmental conditions and the concrete (type of binder) for a particular concrete. α varies
between 0.5 and 0.85.

In some models, cf. Maage & Helland (1991) and Poulsen (1995), a separate expression is
used to estimate the Daex at the time of exposure from laboratory test data
β
⎛t ⎞
D aex = D p ⎜ ex ⎟ [5.3.2c]
⎝ t ⎠

where Dp is a “potential diffusion coefficient” determined from a test method at different


ages. The “aging exponent” α is then the sum of two parts β and γ. Τhe aging exponent β is
due to concrete aging and the aging exponent γ when exposed to sea water is possibly
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 36 of 112
WP4 REPORT – Modelling of Chloride Ingress

explained by ion exchange blocking the concrete surface, Maage & Helland (1991),
Mohammed et al (2002).

5.4.3 Required Input data


Required input data is the apparent diffusion coefficient with its time-dependency Da(t) and
the constant surface chloride content Csa.

5.4.4 Input relation(s) to test methods


The apparent diffusion coefficient Daex at the time of exposure can be estimated from a short
term immersion test. The exponent α, however, must be determined from field data from
appropriate environmental conditions and similar binder types.

5.4.5 Required boundary conditions


The required boundary conditions are simple, the future achieved response by the particular
concrete in the particular environment.

5.4.6 Model(s) for environmental actions


Environmental actions are translated into a surface chloride content Csa mainly from exposure
tests in the true environment, for the concrete that is going to be used. Some models for Csa
are available, i.e. Lindvall (2003), where Csa can be estimated from the concrete composition
(type and amount of binder, w/c etc.), and various environmental parameters, such as sea
water temperature and salinity, distance to sea level, distance to road surfaces etc.

5.4.7 Output data


The prediction gives a chloride profile like the erfc-solution for each exposure time.

5.4.8 Openness/Practicality
The model is simple and fully transparent for any user.

5.4.9 Interpretation of real behaviour


Today it is obvious from numerous field observations that empirical models require a time-
dependent achieved diffusion coefficient Da.

5.4.10 Advantages
The main advantage is the simplicity in performing the predictions and that there is a
possibility to estimate the input data from testing.

5.4.11 Limitations & Drawbacks


Besides the general limitations and drawback for empirical models, see section 5.9, the main
limitation and drawback is the lack of understanding of why the apparent diffusion
coefficient is time-dependent, besides the obvious effect of densification of concrete during
hardening. This lack of understanding gives a large uncertainty in the predictions, since there
is a risk that the time-dependency does not continue to follow empirical equations like
Equation [5.3.2b], e.g. the apparent diffusion coefficient might stop to decrease after some
time.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 37 of 112
WP4 REPORT – Modelling of Chloride Ingress

5.4.12 References
Maage & Helland (1991), Bamforth (1993, 2004), Maage et al (1995), Maage et al (1999),
Mangat (1994), Mohammed et al (2002), Bjergovic (1994), Poulsen (1995), Lee &
Christholm (2000)
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 38 of 112
WP4 REPORT – Modelling of Chloride Ingress

5.5 Models A1c: ERFC-model, D(t) & Csa


(The DuraCrete model)

Time-dependent “true” chloride diffusion coefficients determined from tests started to be


used in the late 1990’s, mainly in the DuraCrete project. By including a number of
“correction factors” correlation with field data was ensured. The uncertainties of all
parameters are quantified and used as input into the model.

5.5.1 Principle
The principle of this model is:

1) Determine a chloride diffusion coefficient for the particular concrete with a short term test
method at a certain age, i.e. 28 days.
2) Calculate an “effective diffusion coefficient” after a certain time of exposure by correcting
the determined diffusion coefficient for the effect of test method, curing, environment and
aging during exposure.
3) A (constant) “apparent surface chloride content” Csa, some of the correction factors and
the exponent in the time-dependency of the diffusion coefficient are estimated from field
exposure of similar concretes in similar environments.
4) Predict the chloride ingress after a certain exposure time, the chloride profile, from the
erfc-solution to Fick’s 2nd law, with Csa and the “effective diffusion coefficient” at that
age.

5.5.2 Mathematics
The original DuraCrete model could be expressed as a model for chloride ingress utilizing
these equations

⎛ x ⎞
C ( x, t ) = C sa ⋅ erfc⎜ ⎟ [5.5.2a]
⎜ 2 D ⋅t ⎟
⎝ e ⎠
n
⎛t ⎞
De = k t ⋅ DRCM ,0 ⋅ k e ⋅ k c ⋅ ⎜ 0 ⎟ [5.5.2b]
⎝t ⎠

- DRCM,0 is the chloride migration coefficient under defined compaction, curing and
environmental conditions, measured at time t0 [m2/s];
- n is an exponent which gives the time-dependency of the effective diffusion coefficient [-];
- kt is a factor which transfers the measured chloride migration coefficient DRCM,0 into a
chloride diffusion coefficient D0 (=kt·DRCM,0) [-];
- ke is a factor which considers the influence of environment on D0 [-];
- kc is a factor which considers the influence of curing on D0 [-];
- Csa is the surface chloride level in [wt % Cl-/binder];
- t is the exposure period [years]; and
- t0 is the reference period [years], in this case t0 = 28 days
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 39 of 112
WP4 REPORT – Modelling of Chloride Ingress

Later, a modification of the DuraCrete model has been done, Gehlen (2000), by “excluding”
a “convection zone” and calculating the chloride penetration from a “peak” in the profile at a
certain depth Δx, cf. Figure 1.2,

⎛ x − Δx ⎞
C ( x, t ) = C sa ,Δx ⋅ erfc⎜ ⎟ [5.5.2c]
⎜ 2 D ⋅t ⎟
⎝ e ⎠

where Δx is in the order of 6-11 mm. This of course means that the modified model only is
relevant for long exposure times where this convection zone has been built up and remained
more or less constant for a long period of time. The time required to build up such a zone,
and the effect of the resistance of the zone, is assumed to be insignificant compared to the
total exposure time and the resistance of the rest of the concrete cover.

5.5.3 Required Input data


The model has at least 6 independent input parameters, most of which are determined from
field data from similar concretes in similar environments.

5.5.4 Input relation(s) to test methods


The migration coefficient DRCM,0 is a direct test result from a test method. Parameters n, kt
and kc may be estimated from the same laboratory test method.

5.5.5 Required boundary conditions


The required boundary conditions are simple, the future achieved response by the particular
concrete in the particular environment.

5.5 Model(s) for environmental actions


Environmental actions are translated into a surface chloride content Csa mainly from exposure
tests in the true environment, for the concrete that is going to be used.

Some models for Csa are available, i.e. Lindvall [2003] originating from the DuraCrete
project, where Csa can be estimated from the concrete composition (type and amount of
binder, w/c etc.), and various environmental parameters, such as sea water temperature and
salinity, distance to sea level, distance to road surfaces etc.

5.5.7 Output data


The prediction gives a chloride profile like the erfc-solution for each exposure time.

5.5.8 Openness/Practicality
The model is simple and fully transparent for any user. It is very straight-forward, but there
are some questions on the definitions of some of the parameters, cf. Section 3.8.

5.5.9 Interpretation of real behaviour


Since the factors are quantified from available field data the model gives prediction results
that coincide with real behaviour. However, the scatter in available data for that
quantification was extremely large.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 40 of 112
WP4 REPORT – Modelling of Chloride Ingress

5.6 Advantages
The model is simple and very straight-forward with input parameters quantified from separate
tests or field exposure data from similar conditions. The model parameters can also be
updated in a simple way when new and better field data is available.

5.6.11 Limitations & Drawbacks


Besides the general limitations and drawback for empirical models, see section 5.9, the
mathematics of the model could be questioned. Equation [5.5.2a] should have the apparent
diffusion coefficient Da, instead of the effective diffusion coefficient De, or correction factors
according to Equation [3.8.2e]. If not, those correction factors must be included in the
environmental factor kt. However, the correction factor in Equation [3.8.2e] includes the
exponent n and considers the difference between age and exposure time. Additionally, if the
time t in Equation [5.5.2b] is the exposure time, the effective diffusion coefficient is not
defined at the start of the exposure.

The model must be revised to correctly consider the time-dependency of the diffusion
coefficient.

5.5.12 References
Alisa et al (1998), Gehlen & Ludwig (1999), Siemes et al (1999), Gehlen (2000), Visser et al
(2002), Gulikers (2004).
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 41 of 112
WP4 REPORT – Modelling of Chloride Ingress

5.6 Models A1d: ERFC-model, Da(t) & Csa(t)


(The False erfc-model)

Some observations in the last few years indicate that the surface chloride content is time-
dependent, also for submerged concrete! This model is the simplest empirical model to
consider that time-dependency. The model, however, is purely empirical and not the correct
mathematical solution to Fick’s 2nd law.

5.6.1 Principle
The principle of this model is:

1) Estimate the time-dependency of the “apparent diffusion coefficient” Da(t) and the time-
dependent “apparent surface chloride content” Csa(t), from field exposure tests.

2) Predict the chloride ingress after a certain exposure time t, the chloride profile, from the
erfc-solution to Fick’s 2nd law, with constant parameters Da(t) and Csa(t).

5.6.2 Mathematics
The model is based on three simple equations

⎛ x ⎞
C ( x, t ) = C sa (t ) ⋅ erfc⎜ ⎟ [5.6.2a]
⎜ 2 D (t ) ⋅ t ⎟
⎝ a ⎠
α
⎛t ⎞
Da = Daex ⎜ ex ⎟ [3.8.1], [5.3.2b] & [5.6.2b]
⎝ t ⎠

C sa (t ) = A ⋅ ln(Δt ini + t − t exp ) + B [5.6.2c]

where t is the age, texp is the age at exposure and Δtini is an initial binding period (14/365
years). A and B are regression parameters.

5.6.3 Required Input data


Required input data are the apparent diffusion coefficient with its time-dependency Da(t) and
the time-dependent surface chloride content Csa(t).

5.6.4 Input relation(s) to test methods


The apparent diffusion coefficient Daex at the time of exposure can be estimated from a short
term immersion test. The exponent α, however, must be determined from field data from
appropriate environmental conditions.

5.6.5 Required boundary conditions


The required boundary conditions are simple, the future achieved response by the particular
concrete in the particular environment.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 42 of 112
WP4 REPORT – Modelling of Chloride Ingress

5.6.6 Model(s) for environmental actions


Environmental actions are translated into a surface chloride content Csa mainly from exposure
tests in the true environment, for the concrete that is going to be used. Models for time-
dependent surface chloride contents are available to some extent, from the Hetek-project, see
section 5.7.

5.6.7 Output data


The prediction gives a chloride profile like the erfc-solution for each exposure time.

5.6.8 Openness/Practicality
The model is very simple and fully transparent for any user. It is very straight-forward, but
there are some questions on the relevance of predictions beyond observation data.

5.6.9 Interpretation of real behaviour


Since the model is based on observations for a particular concrete in a particular environment
the model gives the real behaviour within the observation time. Beyond that, the model seems
to underestimate chloride ingress!

5.6.10 Advantages
The model is very simple and very straight-forward with input parameters quantified from
field exposure data from true conditions. The model parameters can also be updated in a
simple way when more field data is available from longer exposure times. It is a true
“empirical” model!

5.6.11 Limitations & Drawbacks


Besides the general limitations and drawback for empirical models, see section 5.9, one
strong limitation is the lack of mathematical correctness, since the equation [5.6.2a] is not the
true solutions to Fick’s 2nd law with the boundary conditions in Equation [5.6.2c]! That gives
the underestimation of chloride ingress.

5.6.12 References
Nilsson (2002)
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 43 of 112
WP4 REPORT – Modelling of Chloride Ingress

5.7 Models A2: Analytical solutions, DF2(t) & Csa(t)


(The Mejlbro-Poulsen & Hetek model)

Some observations in the last few years indicate that the surface chloride content is time-
dependent, also for submerged concrete! With those boundary conditions, the erfc-equation is
no longer a correct solution to Fick’s 2nd law. A completely new equation must be used.

5.7.1 Principle
The principle of this model is:
1) Estimate the time-dependency of the “apparent diffusion coefficient” Da(t) and the
time-dependent “apparent surface chloride content” Csa(t), from field exposure tests.
2) Predict the chloride ingress after a certain exposure time t, the chloride profile, from
the Mejlbro-Poulsen-solution to Fick’s 2nd law, with time-dependent parameters D(t)
and Csa(t).

5.7.2 Mathematics
Chloride ingress is predicted from this set of equations
⎛ x ⎞
C ( x, t ) = C sa (t exp , t ) ⋅ Ψ p ⎜ ⎟ [5.7.2a]
⎜ 2 D (t ) ⋅ t ⎟
⎝ a ⎠
where
α
⎛ t ex ⎞
Da (t ) = Daex ⎜ ⎟ [5.7.2b]
⎝ t ⎠
and
α p
⎛ ⎛ t exp ⎞ ⎞⎟

C sa (t exp , t ) = C i + S × (t − t exp ) × Daex × ⎜⎜ ⎟ [5.7.2c]
⎜ ⎝ t ⎟⎠ ⎟
⎝ ⎠
where the exponent p depends on how fast Csa(texp,t) increases with time, i.e. mainly on the
type of binder and the environment, cf. Figure 5.7.1.
1.0
p = 0.05

0.8
Surface chloride content

0.6

p =1
0.4

0.2

0.0
0 10 20 30 40 50 60 70 80 90 100
Time of exposure [years]

Fig. 5.7.1 An example of the time-dependency of Csa(texp, t) according to Equation [5.7.2b]


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 44 of 112
WP4 REPORT – Modelling of Chloride Ingress

Examples of the Mejlbro-Poulsen solutions to Fick’s 2nd law with time-dependent diffusion
coefficient and surface chloride content are shown in Figure 5.7.2.

Ψ p (z ) for -0.4 ≤p ≤1; Δ p = 0.2


1.0

0.9

0.8

0.7

0.6
p = -0.4
0.5

0.4

0.3
p = 1.0
0.2

0.1

0.0
0.0 0.5 1.0 1.5 2.0
Abscissa: z

Fig. 5.7.2 Graphs of typical Mejlbro-Poulsen Ψπ-functions.

For p=0 (equal to Csa being constant) the functions are identical to the erfc-equation.

5.7.3 Required Input data


Parameters texp, Daex, α and p are required to make a prediction of chloride ingress. In order to
make the input data physically understandable, the input can as well be estimates of C1, C100,
D1 and D100.

5.7.4 Input relation(s) to test methods


The parameters Cspex and Dpex can be quantified by an immersion test and translated into
values of C1 and D1.

5.7.5 Required boundary conditions


Field observations must be used to quantify the time-dependency of the surface chloride
content Csa(t). For concretes and environments close to the ones already studied, excellent
models are available.

5.7.6 Model(s) for environmental actions


A set of simple models are derived to translate environmental actions into efficiency factors
influencing the surface chloride content and the diffusion coefficient, considering the
concrete composition and the exposure conditions.

5.7.7 Output data


The prediction gives a chloride profile similar to the erfc-solution for each exposure time.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 45 of 112
WP4 REPORT – Modelling of Chloride Ingress

5.7.8 Openness/Practicality
All details of the model have been published. The mathematics involved is not simple to
quantify, however, and software must be made commercially available. Even so, for ordinary
engineers it must be regarded as a ”black box”.

5.7.9 Interpretation of real behaviour


Since the parameters are quantified from available field data from one exposure site the
model gives prediction results that coincide with real behaviour, limited to that site. However,
the parameters can easily be updated if good field data are available from other environments.

5.7.10 Advantages
One important advantage is that the model is the true solution to Fick’s 2nd law with time-
dependent diffusion coefficient and time-dependent surface chloride content. Another
advantage is the models for estimating all the parameters from the concrete composition and
the environmental conditions.

5.7.11 Limitations & Drawbacks


The limitations are very well described by the model developpers: “The Mejlbro-Poulsen
Model do not model the actual physical and chemical processes involved in the chloride
ingress into concrete. The aim of the model is to describe the result of these physical and
chemical processes of this transport, i.e. the chloride profiles. The model is dependent of
good measurements from the environment and the concrete type in question. The
measurements must represent a fairly long time of exposure to improve the predictions. These
drawbacks will be overcomed with time when the experience grow. The model is very
sensitive to the quality of the measurements therefore many of the existing measurements
must regarded as unsuitable for prediction purposes with this model” (Frederiksen et al
(1997)).

An obvious drawback with the model is the complicated mathematics with two different
diffusion coefficients DF2 and Da, both being time-dependent.

5.7.12 References
Mejlbro (1996), Poulsen (1996), Frederiksen et al (1997).
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 46 of 112
WP4 REPORT – Modelling of Chloride Ingress

5.8 Models A3a and A3b: Numerical, DF2(t) & Csa(t)


(NIST, LEO, Life-365, etc.))

Solutions to Fick’s 2nd law with any time-dependency of the diffusion coefficients and the
surface chloride contents can easily be found numerically. The most elaborated one is Life-
365.

5.8.1 Principle
The principle of this model is:

1) Estimate the time-dependency of the diffusion coefficient in Fick’s 2nd law DF2(t) and the,
possibly time-dependent, “apparent surface chloride content” Csa(t), from a chloride
binding isotherm (models A3a) or from field exposure tests (models A3b).

2) Predict the chloride ingress after a certain exposure time t, the chloride profile, from
numerical solutions to Fick’s 2nd law, with time-dependent parameters D(t) and Csa(t).

5.8.2 Mathematics
nd
Fick’s 2 law is used as a pure mass balance equation, the diffusion coefficient DF2(t)
describing the flux of chlorides with the total content of chloride as the driving potential.

∂C ∂ 2C
= DF 2 ⋅ [5.8.2a]
∂t ∂x 2

The diffusion coefficient is time-dependent and temperature dependent

⎛ t ref ⎞
m
⎡U ⎛ 1 1 ⎞⎤
DF 2 (t , T ) = Dref ⋅ ⎜⎜ ⎟⎟ ⋅ exp ⎢ ⋅⎜ − ⎟⎥ [5.8.2b]
⎜T ⎟
⎝ t ⎠ ⎢⎣ R ⎝ ref T ⎠⎥⎦

The boundary conditions are described as Csa(t), in Life-365 increasing at a constant rate up
to a certain maximum level after a defined time. Any other description could be used, in
principle. The time dependency according to equation [5.8.2b] is used in Life-365 for a
limited exposure time only, up to some 25 years. After that, the diffusion coefficient is treated
as a constant.

Solutions to the mass balance equation with these boundary conditions are found numerical
with finite difference methods.

5.8.3 Required Input data


Required input data is the diffusion coefficient in Fick’s 2nd law, with its time-dependency
DF2(t) and the time-dependency of the surface chloride content Csa(t).

5.8.4 Input relation(s) to test methods


Part of the time-dependency of the diffusion coefficient may be derived from laboratory test.
The environmental effect, however, must be found from observations.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 47 of 112
WP4 REPORT – Modelling of Chloride Ingress

5.8.5 Required boundary conditions


Field observations must be used to quantify the time-dependency of the surface chloride
content Csa(t). For concretes and environments close to the ones already studied, reasonable
estimates should be possible. In Life-365 the Csa-values are independent of the concrete
composition.

5.8.6 Model(s) for environmental actions


The boundary conditions, the Csa-values, are completely derived from observations. Field
data is translated into a maximum surface chloride content, with a certain rate of
accumulation during the early exposure.

5.8.12 Output data


The prediction gives a chloride profile similar to the erfc-solution for each exposure time.

5.8.8 Openness/Practicality
Life-365 is widely distributed and tested and a manual is published that describes the details
of the software and the background for the selection of default input data. The user can freely
change all input data, if better data is available. Life-365 could, consequently, be regarded as
a mathematical tool. That could be used for various cases by an educated user.
Similar models can easily be built by using finite difference methods for any set of
assumptions.

5.8.9 Interpretation of real behaviour


Comparisons have been made with laboratory and field data, mainly in quantification of the
input data, the parameters of the diffusion coefficient and the surface chloride content.

5.8.10 Advantages
The main advantage with the Life-365 model is its user-friendliness, especially the data-base
covering environmental conditions around North America.

5.8.11 Limitations & Drawbacks


Besides the general limitations and drawback for empirical models, see section 5.9, there are
uncertainties on the definitions and quantification of the time-dependency of the diffusion
coefficient. The exponent m in Equation [5.8.2b] is mostly quantified from observations of
the time-dependency of the apparent diffusion coefficient, but in other cases from
measurements on old concrete, not exposed to chloride. In Equation [5.8.2a] and [5.8.2b] the
diffusion coefficient is not the apparent one. Consequently, the time-dependency should be
different, cf. Section 3.8. This must be better clarified and defined.

5.8.12 References
Bentz & Thomas (1999), Petre-Lazar et al (2000)
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 48 of 112
WP4 REPORT – Modelling of Chloride Ingress

5.9 Conclusions on models based on Fick’s 2nd law


General conclusions on empirical models are

1. The effect of the other ions in the pore system is completely neglected in empirical
models solving Fick’s 2nd law.
2. The meaning of regression parameters in empirical models is not always easy to
understand. They must be very clearly defined.
3. The background for a number of the assumptions made in empirical models could be
questioned. This is especially important for the continuous time-dependency of the
diffusion coefficients and the surface chloride contents.
4. Empirical models need huge data bases: a Cs for every concrete X, every road Y and
every ”water” Z!

Conclusions 1-3 give rise to questions on whether empirical models can be used at all for
predictions beyond where data exists. Conclusion 4 gives rise to questions on whether
empirical models will ever be practical to use for a new structure, made of a new concrete
exposed in a new environment!
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 49 of 112
WP4 REPORT – Modelling of Chloride Ingress

6 CHLORIDE INGRESS MODELS BASED ON


FLUX EQUATIONS
6.1 General
In “physical models” all physical and (electro)chemical processes are described as
scientifically correct as possible. True physical models use independently determined input
data and no curve fitting to exposure data. Instead, field exposure data is used to validate
predictions. If the comparison between predicted results and exposure data is not good
enough, the model must be improved, or better data must be determined.

Sophisticated physical prediction models for chloride ingress into concrete contains at least
two mass balances, and several relations, for chloride and for water, see Figure 3.

vapour RH v
liquid
water w
free Cf
chloride

bound Cb
chloride

Figure 6.1.1 The mass balance and interaction of chloride and water, in different phases.

The decisive concepts in physical models are concepts in chloride transport and in chloride
binding.

Required input parameters for chloride transport in physical models are several. The diffusion
coefficients should be described at least as functions of temperature, moisture content, degree
of hydration, depth from the surface etc. The description of the convection term requires the
liquid water flow to be described.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 50 of 112
WP4 REPORT – Modelling of Chloride Ingress

6.3 Boundary conditions concepts in physical models


The description of boundary conditions in physical models are much more complicated, since
these models require boundary conditions in terms of chloride, temperature, humidity
conditions at the very surface of the concrete: C(x=0, t), W(x=0, t), T(x=0,t).

An obvious question is: Who’s “responsibility”? Material researchers that develop


sophisticated physical models will find no use of them if quantitative boundary conditions are
not available. For the submerged zone of marine structures it is fairly simple, but for all other
environments it is extremely complicated. The surface conditions at a particular concrete
surface must be regarded as dependent on the location, orientation, distance, time etc.

We have very little data where the same concrete was exposed in two locations. Every
researcher has been exposing his concretes in his own “back yard” with no possibilities to
compare data with another researcher. Additionally we usually miss proper documentation of
the exposure environment. A “splash zone” is not equal to a “splash zone” somewhere else!

A possible approach to make information on boundary conditions for physical models


available in the future is to describe the environmental actions in steps from meteorological
data up to the surface:

• Regional Climate
• Local Climate
• Location of the structure
• Distance from the source of chloride to the concrete surface
• Orientation of the surface

This approach has been analysed for the road environmental actions, cf. Figure 6.2.1. On the
regional scale the macro climatic actions are the ones without a road at all! On the meso
scale, the environmental actions are the one from the road surface, without even considering
the presence of a concrete structure! At the micro scale the effect of the location, the size and
the shape of the concrete structure is considered to give the actual environmental actions at a
particular concrete surface.

Macro Meso (Road) Micro (Surface)


Climate Climate Climate
Air Temperature Rain water splash Equivalent air
Air Humidity Salt water splash temperature
Ground Temperature Rain water fog Surface RH
Radiation Salt water fog Wetness
Rain Wind Salt concentration

Traffic intensity Distance Response by


and speed Height the structure
Road geometry Orientation
Rain shelter Shape
T(x,z.t)
Air streams Surface RH(x,z,t)
Frequency roughness Cl-(x,z,t)
of de-icing

Fig. 6.2.1 An approach to stepwise quantify the environmental actions at a concrete surface
in a road environment, starting from meteorological data.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 51 of 112
WP4 REPORT – Modelling of Chloride Ingress

Without the proper boundary conditions available, we are limited to observations of the
response by concrete in various environments and during various conditions. We do have
some information on the comparison between marine and road climates, cf. Figure 6.2.2.

0.7 0.6
3-50 U1 H4 U1
w /B=0.50, 5%Si; 2 years w /B=0.40, 5%Si; 2 years
0.6 3-50 P1 H4 P1
0.5
3-50 A1 H4 A1
Cl [wt-% of concrete]

Cl [wt-% of concrete]
0.5
207 BK1 0.4 206 DK6
0.4
0.3
0.3
0.2
0.2

0.1 0.1 Rv
Rv4
0 0
0 10 20 30 40 0 10 20 30
Depth [mm] Depth [mm]

0.4
2-75 488d Subm.
Chloride [wt-% of sample]

2-75 488d Splash


0.3
2-75 488d Atm.
236CK6 600d Road
0.2 Träslöv

0.1

Rv40
0.0
0 10 20 30 40 50 60
Depth [mm]

Figure 6.2.2. Chloride profiles for a SRPC concrete with w/b of 0.50 (top left), 0.40 (top
right) and 0.75 (bottom) exposed to various chloride environments, a marine
submerged (upper curves) and atmospheric and road environment (lower
curves).

The build-up of chloride profiles during one season is shown in Figure 6.2.3 for the same
concrete as in Figure 6.2.2 (bottom). The two figures show the response by the vertical and
horizontal surfaces respectively.
0.25
0.25
236 D (3-6)F Feb Feb
0.20 236 D (3-6)Ö
0.20 March March
Cl % Per Sample
Cl [% per Sample]

April 0.15 April


0.15
May May
0.10 Oct 0.10 Oct

0.05 0.05

0.00 0.00
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Depth (mm) Depth (mm)

Figure 6.2.3. Chloride build-up and washout during the first winter and summer of the vertical
(left) and horizontal (right) surface of a concrete with w/c=0.75 in a road
climate.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 52 of 112
WP4 REPORT – Modelling of Chloride Ingress

Even though the chloride profiles have a certain uncertainty, some ±20 %, which makes
comparisons difficult, some indications could be pointed out. Some differences are clearly
larger than the uncertainty. During the winter the chloride content in the surface near region
is very large. The build-up seems to be more rapid for the vertical surface, but later during the
winter, the horizontal surface gets a larger depth of penetration. This could be due to a longer
time-of-wetness for horizontal surfaces, promoting the diffusion and convection of chloride
further into the concrete. On the other hand, the wash-out is also more significant for the
horizontal surface, more or less for the same reason. After the first winter and summer the
chloride profiles are similar for the two orientations, in spite of the differences during the
previous winter.

In the future, that kind of processes must be described if we want to use physical prediction
models for complicated environments. For the time being, these difficulties favour the use of
empirical models, because of the simple description of the boundary conditions, a Cs-value.
That Cs for every individual concrete could be translated into an “equivalent” chloride
concentration at the concrete surface. Then that could be used as boundary conditions for a
new concrete, without having to perform an exposure program.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 53 of 112
WP4 REPORT – Modelling of Chloride Ingress

6.3 Models B1a: Models based on Fick’s 1st law of diffusion, no


convection
(ClinConc, etc.)

One of the most advanced models for predicting chloride penetration into concrete, ClinConc,
was presented by Tang & Nilsson [1994] and Tang [1996]. By using a FDM numerical
approach, most of the factors involved in chloride penetration are considered in a relevant
and scientific way.

6.3.1 Principle
The chloride flux is described with the chloride diffusion coefficient in Fick’s 1st law DF1,
calculated from a separate migration test at a certain age. The effect of densification of
concrete on the diffusion coefficient is included up to a certain age, i.e. six months for
Portland cement concrete.

Chloride binding is described by a binding isotherm that is a function of pH and temperature.

Leaching of alkalis is described as a pure diffusion of hydroxides out of the concrete and the
pH- profile is part of the prediction in every time-step. Only the alkali hydroxides are dealt
with, e.g. pH remains larger than 12.5.

The mass balance equations for chlorides and hydroxides are solved by using separate terms
for chloride diffusion and for chloride binding. In every time step the fluxes of chloride is
calculated at each depth with Fick’s 1st law. From differences in fluxes the change in total
chloride content is calculated at each depth. By using the binding isotherm, the change in
total chloride content is divided into free and bound chloride.

The chloride ingress is shown as profiles of free and total content of chloride.

Both semi-infinite cases and cases with limited thicknesses can be treated.

6.3.2 Mathematics
The diffusion coefficient Di,j in Fick’s 1st law is a function of temperature, age and depth:

Di , j (Cl ) = D0 T0 ( )( )
f D Tj g t j f ( xi ) [6.3.2a]

The age dependency of the diffusion coefficient is calculated from

⎧⎛ t ⎞ β t

⎪ 0 t < t0
g (t ) = ⎨⎜⎝ t ⎟⎠ [6.3.2b]
⎪ 1 t ≥ t0

The total amount of chloride is split into free (in a moisture content Ws) and bound
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 54 of 112
WP4 REPORT – Modelling of Chloride Ingress

⎡ kg ⎤
C i , j ( total) = ci , j (Ws ) i , j + C i , j (bound) ⎢ 3 Cl ⎥ [6.3.2c]
⎣ m concrete ⎦

where the binding isotherm is expressed by

⎛ [OH ]i , j −1 ⎞⎟ Eb ⎛⎜ 1 1 ⎞
α OH ⎜ 1− − ⎟
⎜ [OH ]initial , j −1 ⎟⎠ R ⎜⎝ T j T0 ⎟ fb B
C i , j (bound ) = e ⎝
e ⎠
Wi , j (gel) ci , j [6.3.2d]
1000

6.3.2 Required Input data


Input data to the model is the complete concrete composition, the diffusion coefficient with
its time-dependency and the chloride binding isotherm as a function of pH and temperature.
The basic diffusion coefficient is the “true” chloride diffusion coefficient DF1, evaluated from
a migration test method. Additionally, the effect of depth from a cast surface can be taken
into account. The effect of age is described with a time-dependency of the diffusion
coefficient up to a certain age.
For a new Portland cement concrete only one parameter has to be determined, the diffusion
coefficient for the particular concrete at a certain age. The other parameters in the model
follow from the mix composition and the exposure conditions. For a new binder, the
parameters of the binding isotherm must be given.

6.3.4 Input relation(s) to test methods


The diffusion coefficient at a specified age is determined in an independent, short-term test
with no curve-fitting is involved.

6.3.5 Required boundary conditions


The boundary conditions can described as sinusoidal annual variation of concentration of
chloride and water temperature in the surrounding chloride solution.

6.3.6 Model(s) for environmental actions


In later version, the chloride concentration at the surface can be varied in time stepwise.

6.3.7 Output data


The predictions give three profiles: free and total concentration of chloride and concentration
of alkalis. They are shown graphically and in tables. The profiles do not always have an erfc-
shape, depending on temperature and leaching conditions.

6.3.8 Openness/Practicality
The software is not yet commercially available but spread among researchers. All details,
however, are thoroughly published.

The user-friendliness makes the model extremely practical. For a new concrete only one test
has to performed, to determine the diffusion coefficient at a certain age.

6.3.9 Interpretation of real behaviour


A number of comparisons have been made with laboratory and field data. The predictions for
a number of concretes submerged in sea water was once made with the independently
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 55 of 112
WP4 REPORT – Modelling of Chloride Ingress

determined diffusion coefficients The predictions coincided very well with the field data from
up to two years of exposure. Then, for some concretes, the measured chloride profiles had
larger surface chloride contents than the predictions. A time-dependency of the chloride
binding had to be empirically included to obtain a better correlation. This time-dependency of
chloride binding, however, remains to be explained and independently quantified.

6.3.10 Advantages
The most significant advantage is the small number of input parameters and that the only one
that has to be determined, can be quantified in a separate, independent test. Another
advantage is that the effect of temperature variations and leaching can be considered in a
physical way.

Recently, ClinConc has been expressed in an engineering way for convenient applications
(Tang 2005).

6.3.11 Limitations & Drawbacks


The physical part of the model can only treat cases with saturated concrete exposed in
submerged conditions. In later versions empirical parts have been added that combines
stepwise changed chloride concentrations in the surrounding environment. The concrete,
however, still is regarded as saturated.

The exclusion of the effects of all ions on the chloride transport process gives rise to
questions on the ability to describe future ingress in a correct way.

6.3.12 References
Tang (1996, 2005)
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 56 of 112
WP4 REPORT – Modelling of Chloride Ingress

6.4 Models B1b: Models based on Fick’s 1st law of diffusion,


with convection
(Imperial, HetekConv, ConFlux, THI EdF, LERM, Meijers, EPFL etc.)

A number of similar physical models are available where a convection term has been added,
to be able to predict chloride ingress in a number of applications where moisture flow and
wetting and drying play a dominating role. These models have different simplifications and
assumptions, but they are here described together. Examples of details are taken from the
Hetek Convection model, partly because it is fairly elaborated and partly because it is best
known by the author(s).

6.4.1 Principle
The convection of chloride is governed by the moisture flow and the changes in moisture
contents will change the chloride concentrations. Consequently, the moisture distributions
must be predicted by solving the mass balance equation for moisture. The moisture flow must
be separated in two different fluxes, only one carrying the ions.

The mass balance equation for chloride, and the separation into free and bound chloride, can
be described by taking the moisture content into account, since the free chlorides are limited
to the water filled parts of the pore system.

The two mass balance equations are coupled and must be solved together. The main part of
the solution is to describe the flows of moisture and chloride. The initial conditions are trivial
but the boundary conditions must be, until better data are available, a simplification of the
true environmental conditions at the concrete surface.

6.4.2 Mathematics
The moisture flow is divided into vapour flow and liquid flow, i.e. with vapour content and
pore water pressure as driving potentials, respectively

∂v ∂P
q w = q v + ql = −δ v ( RH ) − k P ( w) w [6.4.2a]
∂x ∂x

The chloride flux is described by Fick’s 1st law, in the water filled part of the pore system,
e.g. the diffusion coefficient is a function of the moisture content, and one term giving the
convection of chloride with the liquid water flow

∂c 'f
q Cl = − DCl ( w, T , c 'f ) + q l c 'f [6.4.2b]
∂x

A number of relationships must be expressed mathematically, i.e. by considering that the


moisture sorption depends on the chloride concentration and by describing the separation of
chloride into free and bound chloride, possibly with a non-linear binding isotherm, including
the moisture content in the pore system.

This is all done differently in the available models.


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 57 of 112
WP4 REPORT – Modelling of Chloride Ingress

6.4.2 Required Input data


Required input data for this kind of physical models, at least the most sophisticated ones, is
tremendous! The chloride and moisture flow coefficients must be given as functions of a
number of parameters. A number of relationships between all parameters require a large
number of input parameters to be quantified.

The input data for boundary conditions require the duration of dry and wet periods at the
concrete surface and the temperature and humidity conditions at the very surface during these
periods. Additionally boundary conditions must be described as concentrations of free
chloride in the solution in contact with the concrete surface, when it is present.

6.4.4 Input relation(s) to test methods


The input data for the material properties and relationship can all be determined by test
methods. Those test methods, however, are not yet standardized and in some cases not yet
invented!

As long as this input data is missing to a large extent, physical models can only be used in
very special cases where data exists.

6.4.5 Required boundary conditions


The boundary conditions at the concrete surface must describe the environmental actions as a
function of time in terms of time-varying chloride concentration in a solution in contact with
the surface, alternating with drying conditions.

6.4.6 Model(s) for environmental actions


For non-trivial cases no models exist today that can translate environmental actions in
practice to the required boundary conditions at the concrete surface in physical models.

6.4.7 Output data


Physical models of this kind predict the moisture profiles and the profiles of free and total
concentration of chloride.

6.4.8 Openness/Practicality
Physical models are rarely commercially available or user-friendly at all. In most cases the
only possible user is the developer himself! Additionally, the models must treat so many
important details in a correct way that it is very difficult for anyone else to evaluate the
accuracy and relevance of a physical model. A number of models, however, are very well
documented and all details are thoroughly published.

The large amount of required input data is not easily available but have to be quantified by
laboratory measurements and complicated calculations for every new case. This makes the
general practical application of physical models very limited.

6.4.9 Interpretation of real behaviour


All models “produce” nice results and frequently predictions are fitted to measured data by
adjusting a number of all the input parameters.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 58 of 112
WP4 REPORT – Modelling of Chloride Ingress

However, good data that can be used for comparisons is scarce. Simultaneously measured
moisture and chloride profiles must be made available, where micro climatic data at the
concrete surface and all relevant material properties are well documented.

6.4.10 Advantages
Physical models of this kind are the only ones capable of handling cases where moisture flow
plays a dominant role. Examples are concrete structures with limited thickness that are
exposed to chloride on one side only (tunnels, pool walls, foundation walls), structures
exposed to drying and wetting, sometimes with splash from de-icing salts (road structures) or
sea water (marine splash and atmospheric zones). They are also capable of handling chloride
transport in material combinations.

6.4.11 Limitations & Drawbacks


The tremendous amount of required input data, material properties and environmental
conditions, most of it as functions of the other parameters, makes the practical application
very doubtful.

The relevance of all assumptions and limitation is not very well known or quantified.

The exclusion of the effects of all ions on the chloride transport process gives rise to
questions on the ability to describe future ingress in a correct way.

6.4.12 References
Buenfeld (1995), Nilsson (1997, 2000), Hooton, Thomas et al (199x), McLoughlin I.M
(1997), Petre-Lazar et al (2003), Meijers (2003), Denarié et al (2003).
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 59 of 112
WP4 REPORT – Modelling of Chloride Ingress

6.5 Models B2a, b and c: Physical models based on the


Nernst-Planck flux equation
(MS-Diff, Li & Page, Johannesson, Stadium)

6.5.1 Principle
The fluxes of all ions in the system are predicted by considering how they influence each
other by creating an electrical field in the pore solution. This electrical field must be predicted
simultaneously. The profiles of each specie are then predicted by solving the mass-balance
equations for all ions.

6.5.2 Mathematics
The models must describe the flux of every individual type of ion i, which is done with the
Nernst-Planck equation

⎛ ∂c ∂ ln ai z i F ∂Φ ⎞
qi = − Di ⎜ i + ci + ci ⎟ [3.8] & [6.5.2a]
⎝ ∂x ∂x RT ∂x ⎠

In some of the models a convection term is added to consider the effect of liquid moisture
flow. Some models even include the effect of degradation and micro structural changes that
will increase the diffusion coefficients over time.

The flux equations include the electrical field. This electrical field is created by the species of
opposite sign, so that the faster ions will be slowed down and the slower ones will be
accelerated. A general expression for the electrical field, Masi et al (1997), is given by:

I (t ) ∂c ( x , t )
+ ∑ z i Di i
∂Φ RT SF ∂x [6.5.2b]
( x, t ) = − i

∂x F ∑ i i i )
z
i
2
D c ( x , t

where I(t) is the electrical current and S is the cross-section of the material. The Poisson
equation may be used instead of Eq. [6.5.2.b] in order to determine the electrostatic potential
(Marchand et al, 2002):

∂ ²Φ F ⎡ N ⎤
+ ⎢ ∑ z i c i ( x, t ) ⎥ = 0 [6.5.2.c]
∂x ² ε ⎣ ⎦

where N is the total number of ionic species and ε is the dielectric permittivity of the
medium.

For every type of ion the mass-balance equation is solved

∂ci ( x, t ) ∂q ( x, t ) ∂Cb, i ( x, t )
psol =− i − [6.5.2c]
∂t ∂x ∂t
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 60 of 112
WP4 REPORT – Modelling of Chloride Ingress

where the right-hand term gives the binding/interaction between species i and the matrix, e.g.
chloride binding or some kind of reaction or liberation.

6.5.3 Required Input data


Required input data for this kind of physical models, is tremendous, especially if convection
is considered! The ion diffusion coefficients and moisture flow coefficients must be given as
functions of a number of parameters. A number of relationships between all parameters
require a large number of input parameters to be quantified.

The input data for boundary conditions require the duration of dry and wet periods at the
concrete surface and the temperature and humidity conditions at the very surface during these
periods. Additionally boundary conditions must be described as concentrations of ions in the
solution in contact with the concrete surface, when it is present!

Parameters of the binding/interaction term must be quantified separately. Different models


use various ways to consider e.g. chloride binding. Truc (2000) and Li & Page (1998) used a
chloride binding isotherm (models type B2a), Johannesson (2000) described chloride binding
as a special ion-solid interaction process (models type B2b) and Samson et al (1999)
concentrated on the chemical fixation of chloride (models type B2c).

6.5.4 Input relation(s) to test methods


The diffusion coefficients Di for each type of ion cannot be determined in a separate test,
since the flux of one ion depends on what other ions are present and at what concentrations.
Samson et al (1999) overcame this obstacle in a suitable way by determining a “formation
factor” for the pore system with another test method and then simply multiplied the formation
factor with the diffusion coefficient for a type of ion in solution. If this is correct, a test
method for the formation factor is required.

Khitab et al (2004) proposed to measure the chloride diffusion coefficient from a migration
test (LMDC-test.) The other ionic species diffusion coefficients are linked to the chloride one
by the same relation as in an infinitely diluted solution.

6.5.5 Required boundary conditions


The boundary conditions at the concrete surface must describe the environmental actions as a
function of time in terms of time-varying chloride concentration in a solution in contact with
the surface, alternating with drying conditions.

6.5.6 Model(s) for environmental actions


For non-trivial cases no models exist today that can translate environmental actions in
practice to the required boundary conditions at the concrete surface in physical models. In
reality this means that only simple boundary conditions can be treated, such as direct contact
with a salt solution or drying in an atmosphere with a given humidity.

6.5.7 Output data


Physical models of this kind predict the moisture profiles, if convection is considered, and the
profiles of free and total concentration of all ions. An example of a prediction of the ion
profiles and ion fluxes and the electrical potential are shown in Figure 6.5, for a 3 cm thick
concrete specimen in a migration test.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 61 of 112
WP4 REPORT – Modelling of Chloride Ingress

Fig. 6.5 Predicted concentration profiles and fluxes for chloride, potassium, sodium and
hydroxide in a 3 cm thick specimen in a migration test where 12 V potential
difference was applied. The predicted data is shown when chloride penetrated
approximately 1 cm, Truc (2000).

6.5.8 Openness/Practicality
Physical models are rarely commercially available or user-friendly at all. In most cases the
only possible user is the developer himself! Additionally, the models must treat so many
important details in a correct way that it is very difficult for anyone else to evaluate the
accuracy and relevance of a physical model. A number of models, however, are very well
documented and all details are thoroughly published.

The large amount of required input data is not easily available but have to be quantified by
laboratory measurements and complicated calculations for every new case. This makes the
general practical application of physical models very limited.

6.5.9 Interpretation of real behaviour


All models “produce” nice results and frequently predictions are fitted to measured data by
adjusting a number of all the input parameters.

However, good data that can be used for comparisons is scarce. Simultaneously measured
moisture and ion profiles must be made available, where micro climatic data at the concrete
surface and all relevant material properties are well documented.

6.5.10 Advantages
Physical models of this kind are the only ones capable of considering the effects of all ions in
a correct way. This makes them especially suitable for explaining observations in laboratory
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 62 of 112
WP4 REPORT – Modelling of Chloride Ingress

and field tests that cannot be foreseen by simple empirical and physical models. In this way
the confidence in simple models might be significantly improved.

6.5.10 Limitations & Drawbacks


The tremendous amount of required input data, material properties and environmental
conditions, most of it as functions of the other parameters, makes the practical application
very doubtful. Relevant test methods for the diffusion coefficients must be verified.

The interaction/binding processes are still not fully understood and possible to quantify.

6.5.12 References
Li & Page (1998), Samson et al (1999), Truc (2000), Johannesson (2000), Marchand et al
(2002), Khitab et al (2004).
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 63 of 112
WP4 REPORT – Modelling of Chloride Ingress

6.6 Conclusions, models based on flux equations


The main conclusions on the models based on flux equations are

• Physical models, especially those considering all ions and convection at the same
time, give theoretical predictions as close to our present knowledge as possible.

• Physical models may be used to check the relevance of more simple, empirical
models and may be explain some the assumptions that have to made in simple
models. That may improve the confidence in empirical models.

• Physical models need quantification of a number of material parameters and a number


of environmental parameters, each as a function of several parameters,

• The required surface climatic conditions are extremely complicated to model.

• Because of lack of availability, limited user-friendliness and the huge amount of


required input data most physical models will remain research tools, not meant for
practical applications.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 64 of 112
WP4 REPORT – Modelling of Chloride Ingress

7 SHORT TERM SENSITIVITY ANALYSIS


7.1 Introduction
This chapter presents the methodology used to perform the short term (10 year exposure)
sensitivity analysis on models for chloride penetration prediction. We need to:

- define a practical situation


- define the model
- define the input data and their uncertainty (for the considered practical situation)
- define a limit state
- do probabilities (run the model with several set of input data)

We obtain the uncertainty of the final result and the importance factors of the input data.

The models which have been tested are:

- the ERFC model with constant diffusion coefficient and constant surface
concentration,
- the LEO model,
- the Ms-Diff model.

For more information on these models, see chapter 5 and 6.

The work in WP4.2 was mainly performed by X.-S. Nguyen, M. Carcasses, I. Petre-Lazar, G.
Martin at INSA and EDF, respectively.

7.2 Methodology for sensitivity analysis


Usually, the input data for a model are random variables. The environmental parameters are
not very well defined and even if the concrete formulation is well known, its properties vary
in all the construction. Though a deterministic approach is not enough and it’s important to
know the influence of input data variation on output data for each model. So a probabilistic
approach can help to evaluate these variations.

7.2.1 Probabilistic method principle


Generally, we search the probability of a function G(X) (where X is the random variable
vector of input data for a model) to be greater than 0. G(X) = 0 is the limit state function and
separate input data space in two sets: the failure set (G(X) < 0) and the safe set (G(X) > 0).
To resolve this kind of problem, different numerical methods exist as simulation methods (for
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 65 of 112
WP4 REPORT – Modelling of Chloride Ingress

example the Monte-Carlo one) or others like FORM and SORM (respectively First et Second
Order Reliability Method). These methods are available in the probabilistic software
PROBAN, Olesen (1992), developed by Det Norske Veritas. With this software, it is possible
to determine the failure probability (P(G(X)<0), the most probable failure point and the
importance factors, Madsen et al (1986). For each input data, its importance factor measure
the relative influence of the input data variability on the failure probability. This means that
the higher the importance factor for an input data is, the higher the influence of the input data
variability will be on the position in safe or unsafe set.

Refer to probability law of a random variable:


Mathematically, a random variable is defined as a value which depends on the result of a
random test e.g. not previewed in advance with good accuracy. The probability law of a
random variable allows calculating these values. A probability law FX can be characterised by
a repartition function R:

F X : R → [0,1] x → FX ( x ) = P ( X ≤ x ) [7.1]

where X is a random variable, x its realisation and FX is an increasing function, continuous on


the right as lim FX ( x) = 0 and lim FX ( x) = 1 . So it’s possible to calculate the probability for
x → −∞ x → +∞

X to belong to any interval of R :


∀(a, b) ∈ R 2 , a < b, P(a < X ≤ b) = FX (b) − FX (a )
The random variables are classified as a function of their value validity set.
Therefore, a random variable X belongs to a Gaussian distribution law (normal law) if its
− ( x −m )2
1
values are in R and FX ( x) = e 2σ 2
.
σ 2π
The beta distribution law is expressed as follow:
( x − a) r −1 (b − x) t − r −1
FX ( x ) = [7.2]
(b − a) t −1 B(r , t − r )
with
1
B (r , s ) = ∫ t r −1 (1 − t ) s −1 dt [7.3]
0

This last choice allows to correctly model the data defined by “an inferior limit, a most
probable value and a superior limit”. The a and b parameters correspond to the limits and xmax
is the most probable value in the set [a,b]. Hence, the mean value m and the standard
deviation σ of the variable are :

a + 4 x max + b b−a
m= ;σ = [7.4]
6 6

7.2.2 Definition of a practical situation


A practical situation has to be defined by one concrete in one particular environment and for
a defined service life. We have chosen the case of free diffusion of Cl- in a saturated concrete.
The input data for a particular model are based on the experimental data corresponding to this
situation obtained for the defined concrete.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 66 of 112
WP4 REPORT – Modelling of Chloride Ingress

The environment is pure water with NaCl [Cl-] = 18 g/l in the upstream compartment and 0 in
the downstream one.
The simulations are made for the concrete defined in the table 1.
The composition of the initial solution in concrete porosity is 23 mol/m3 of Na+, 156 mol/m3
of K+, 0 mol/m3 of Cl- et 179 mol/m3 of OH-.

Table 7.1 : Characterisation of the concrete used for simulation

No. Description Unit Quantity


3
1 Cement Kg/m 560
Sand Kg/m3 695
Gravel Kg/m3 825
2 Water Litre/m3 224
3 Paste Kg/m3 784
4 Concrete Water porosity % 16
5 Solid density Kg/m3 2710
e / c − p(e / c + 0,32)
Concrete hydration h= = 0,79
5 0,36
coefficient
Powers’ formula
6 Concrete dry density Kg/m3 2304

The concrete effective diffusion coefficient has been measured at 28 day old with the LMDC
test:
Sample 1: De1 = 1.71·10-12 m2/s
Sample 2: De2 = 1.72·10-12 m2/s
Sample 3: De3 = 1.76·10-12 m2/s

The De coefficient is a random variable with a normal distribution law. The mean value is:
(1.71 + 1.72 + 1.76) ⋅10 −12
De , moyen = = 1.73 ⋅10 −12 m 2 / s
3
and its variation coefficient is 1.25%.

7.2.3 Model definition


A model allows predicting chloride penetration as a function of time with input data
concerning concrete properties and information on the environment.
We have studied the sensitivity of three models:

- the ERFC model with constant diffusion coefficient and constant surface
concentration,
- the LEO model,
- the MS-Diff model.

For some models, the input data presents a great variability and used a probabilistic approach
Petre-Lazar (2000). Each parameter is defined by its distribution law. For the measured data,
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 67 of 112
WP4 REPORT – Modelling of Chloride Ingress

we proposed a Gaussian distribution law. For the parameters which are not very well known
but for which the set of variation is defined, we propose a beta distribution law.
In our case, we study the influence of two particular input data which are used in all the
models: the chloride diffusion coefficient and the chloride surface content.

7.2.4 State limit definition


A limit state is required by the probabilistic method (FORM) in order to perform the
sensitivity analysis. This limit state defines the transition between the safe set and the unsafe
set. Nevertheless, the choice of the state limit has a great influence on the sensitivity analysis.
So in our case we proposed a limit state concerning steel corrosion. For example, we define a
critical concentration of chloride which initiate corrosion on the steel surface in concrete even
though the definition of such a value is still discussed.
We have chosen a cover depth of 5 cm from the concrete exposed surface and a chloride
critical concentration expressed by volume of pore solution: Ccrit = 6,5 g Cl-/l. So we consider
that corrosion is initiated at a depth of 5 cm when chloride concentration is greater than Ccrit =
6,5 g/l. Hence, the state limit function can be written as follows:

G = Ccrit - C(x=5cm,t) [7.5]

7.2.5 Analysis execution


To do the probabilistic analysis, we use the Monte-Carlo simulation and the FORM/SORM
method. The Monte-Carlo method is based on a lot of random input data sets. For each set,
the response of the model is calculated and a statistical treatment of whole the results is
applied. The FORM / SORM and Monte-Carlo methods allow calculating the importance
factors.

The FORM/SORM method principle can be represented by the Figure 7.1.

U2 g (S(X(U))) β : fiability index


≡ G(U) = 0 U* : design point
X : random variable
S(X) : system response
β SORM
g(X,S(X)) : state limit function –
U* failure criteria
FORM

U1

Figure 7.1 Geometrical representation of the conception point

The FORM method is used to:


- Evaluate the fiability index β. The design point U* is determined by an optimisation
algorithm :
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 68 of 112
WP4 REPORT – Modelling of Chloride Ingress

U * = Arg min U { 2
}
G (U , S (U )) ≤ 0 [7.6]

- Obtain a n approximate value of the failure probability.

Pf ≈ Φ (− β ) [7.7]

To increase the precision of equation 7.7, the SORM method has been proposed. The idea is
to replace the limit state surface by a quadratic surface which probabilistic content is
analytically known.

n −1 1
Pf ≈ Φ (− β ) Π [7.8]
i =1
1 + βk i

The Monte-Carlo method can be represented by the following figures:

Set of random Numerical


variables model g(S(Xi)) Nf
Pf ≈
q

Unsafe
set

Nf

10k+2 set to
evaluate Pf = 10-k

Figure 7.2 Illustration of the Monte-Carlo method

These methods are available in the PROBAN software.

Figure 7.3 illustrate the realisation of the sensitivity analysis:


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 69 of 112
WP4 REPORT – Modelling of Chloride Ingress

INPUT DATA
Random value set

SOFTWARE
Probabilistic tool : PROBAN
Numerical model
OUTOUT DATA
Range of output data
Importance factors

Figure 7.3 Representation of the sensitivity analysis

The output is a couple of values which characterise the range of the output data. We calculate
the fractiles at 95% and 5%. They are determined by the following relations:

f 95 = m + 1.645σ (superior limit) f 5 = m − 1.645σ (inferior limit)


where m is the mean value and σ the standard deviation of all the obtained output data.

7.3 Application of the methodology for ERF

7.3.1 Definition of the model


This is the straight forward error function:
⎡ ⎛ x ⎞⎤
C ( x, t ) = C surf .⎢1 − erf ⎜ ⎟⎥ [7.9]
⎢⎣ ⎜ 2 D .t ⎟⎥
⎝ a ⎠⎦
where C(x,t) is the chloride concentration in time [g of Cl/l], Csurf is the chloride
concentration in the external solution [g of Cl/l], Da is the apparent diffusion coefficient
[m2/s], t is the exposure time [s] and x is the penetration depth [m].

7.3.2 Input data


Table 7.2 : Input data for ERF

No. Parameter Meaning Statistical distribution


1 Gaussian distribution
Csurf Chloride surface concentration Mean value = 18 g/l
Coefficient of variation = 5%
2 Gaussian distribution
Da Diffusion coefficient Mean value = 4.96·10-12 m2/s
Coefficient of variation = 1.25%
3 x Penetration depth 0.05 m (protection thickness)
4 Ccrit Limit state of corrosion initiation 6.5 g/l
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 70 of 112
WP4 REPORT – Modelling of Chloride Ingress

7.3.3 Results for ERF

Chloride profiles:
22
20
concentration (g/l) 18
16
14
12 free chlorides
10 mean profile
8
6
4
2
0
0 0,01 0,02 0,03 0,04 0,05
depth x (m)
At t = 10 years:

Figure 7.4 : Band of chloride profiles at 10 years for ERF

Importance factors
In our case, we interest in the importance factors before 10 years. The results of importance
factors are presented in the table 3 :
Table 7.3 : Importance factor in time evolution for ERF
Importance factors
t (years)
Csurf Da
3 69.4 30.6
4 80.0 20.0
5 87.2 12.8
6 91.8 8.2
7 94.5 5.5
8 96.1 3.9
9 97.2 2.8
10 97.8 2.2
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 71 of 112
WP4 REPORT – Modelling of Chloride Ingress

100
90
80
percentage (%)

70
60 Csurf
50
40 Da
30
20
10
0
0 5 10
T (years)

Figure 7.5 Importance factor in time evolution for ERF

On the Figure 7.5, we found that the importance factors are time-dependent. In considered
case, the variation of the chloride surface concentration is more important than the one of the
chloride coefficient.

Curve of the limit state function


A limit state defines the transition between the safe set and failure set.

7,50E-12

7,00E-12

6,50E-12 None passivation


zone
D (m2/s)

4 years
6,00E-12

5,50E-12 10 years

5,00E-12
Passivation
4,50E-12
zone
4,00E-12
10 15 20 25 30 35
Csurf (g/l)

Figure 7.6 Curve of the limit state function for ERF


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 72 of 112
WP4 REPORT – Modelling of Chloride Ingress

7.4 Application of the methodology for LEO

7.4.1 Definition of the model


This is the function of the model:
⎡ ⎛ x ⎞⎤
C ( x, t ) = C initiale + (C surf − C initiale ) ⎢1 − erf ⎜ ⎟⎥ [7.10]
⎜ ⎟⎥
⎢⎣ ⎝ 2 α .η .DCl .t ⎠⎦
where C(x,t) is the chloride concentration in time [g/l], Cinitiale is the initial chloride
concentration in the pore solution [g/l], Csurf is the chloride concentration in the external
solution [g/l], DCl is the effective chloride concentration in the pore solution [m2/s], x is the
penetration depth [m], t is the exposure time [s], α is the correction coefficient on the
interaction of ionic flux and η is the correction coefficient on the interaction of ion – matrix.

7.4.2 Input data

Table 7.4 : Input data for LEO


No. Parameter Meaning Statistical distribution
Initial chloride concentration in
1 Cinitiale 0 (g/l)
the pore solution
Gaussian distribution
2 Csurf Chloride surface concentration Mean value =18 (g/l solution)
Coefficient of variation = 5%
Gaussian distribution
Effective diffusion coefficient in
3 D Mean value = 10.81·10-12 (m2/s)
pore solution
Coefficient of variation = 1.25%
4 x Penetration depth 0.05 m (protection thickness)
5 Ccrit Limit state of corrosion initiation 6.5 g/l
6 cement Cement quantity/m3 of concrete 560 kg
7 water Water quantity/m3 of concrete 224 kg
8 paste Paste quantity/m3 of concrete 784 kg
9 p Concrete porosity 0.16
10 h Concrete hydratation coefficient 0.79
11 ρ Density of concrete 2304 kg/m3
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 73 of 112
WP4 REPORT – Modelling of Chloride Ingress

7.4.3 Results for LEO

Chloride profiles
22
20
concentration (g/l) 18
16
14
12 free chlorures
10 mean profile
8
6
4
2
0
0 0,01 0,02 0,03 0,04 0,05
depth x (m)
At t = 10 years :

Figure 7.7 Band of chloride profiles at 10 years for LEO

Importance factors

Table 7.5 : Importance factor in time evolution for LEO

Importance factors
t (years)
Csurf D
4 62.7 37.3
5 72.8 27.2
6 81.5 18.5
7 87.8 12.2
8 92.0 8.0
9 94.6 5.4
10 96.3 3.7
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 74 of 112
WP4 REPORT – Modelling of Chloride Ingress

100
90
80
percentage (%)
70
60 Csurf
50
40 D
30
20
10
0
0 2 4 6 8 10
T (years)

Figure 7.8 Importance factor in time evolution for LEO

As the previous case, the importance factors are time-dependent and the variation of the
chloride surface concentration is more important than the one of the chloride coefficient.

Curve of the limit state function

2.20E-11

2.00E-11
None
passivation zone
1.80E-11 4 years
D (m2/s)

1.60E-11

1.40E-11
10 years
1.20E-11 Passivation
zone

1.00E-11
15 20 25 30 35 40 45
Csurf (g/l)

Figure 7.9 Curve of the limit state function for LEO


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 75 of 112
WP4 REPORT – Modelling of Chloride Ingress

7.5 Application of the methodology for MS-Diff

7.5.1 Definition of the model


The flux of ionic species is described by Nernst Planck equation:
∂ci ( x, t ) ∂C mi , B (ci ( x, t ))
p + (1 − p 0 ) ρ =
∂t ∂t
[7.11]
∂ ⎡ ∂c ( x, t ) c ( x, t ) ∂γ i ( x, t ) F ∂ϕ ( x, t ) ⎤
− ⎢ De ,i i + De , i i + z i De ,i ci ( x, t ) ⎥
∂x ⎣ ∂t γ i ( x, t ) ∂x RT ∂x ⎦

where De.i is the diffusion coefficient of the ionic specie i [m2/s], ci is the concentration of
ionic specie i [mol /m3], γi is the activity coefficient of the ionic specie i, zi is the valence of
the ionic specie i, ϕ is the electric potential [V], F is the constant of Faraday [96480 J/(V x
mol)], R is the constant of perfect gas [8.314 J/(mol x K)], T is the absolute temperature [K],
po is the porosity of concrete, ρ is the density of dry concrete [kg /m3] and Cmi.B is the number
of bound specie divided by the dry material [mol /kg].

In the MsDiff model, we consider four ionic species Na+. K+. OH- et Cl-.

7.5.2 Input data

Table 7.6 : Input data for MsDiff

No. Descrition Unit Value, incertainty


1 Sample thickness m 0.2
2 Exposed sample surface cm2 95
3 Porosity % 16
3
4 Dry volume mass kg/m 2710
5 Freundlich isotherm coefficients
α 0.0064
β 0.33
6 Immersion duration days
7 Time step sec 200.000
8 Space step nodes 100

Gaussian distribution
Chloride diffusion coefficient Mean value = 1.73·10-12 m2/s
9 m2/s
(LMDC test) Coefficient of variation = 1.25
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 76 of 112
WP4 REPORT – Modelling of Chloride Ingress

10 Ratio Dion/DCl-
Na+ 0.05
K+ 0.0735
OH- 2.73
11 Specie concentration in the upstream compartment
Na+ mol/m3 Cl-
K+ mol/m3 0
Gaussian distribution
Cl- mol/m3 Mean value = 18 g/l = 507 mol/m3
Coefficient of variation 5%
OH- mol/m3 0

12 Specie concentration in the interstitial solution and the downstream compartment

Na+ mol/m3 23
3
K+ mol/m 156
3
Cl- mol/m 1
3
OH- mol/m 178
13 External current mA 0

7.5.3 Results for MsDiff

Chloride profiles
At t = 10 years:
22
20
18
concentration (g/l)

16
14
12
free chloride
10
8
6
4
2
0
0 0,01 0,02 0,03 0,04 0,05
depth (m)

Figure 7.10 Band of chloride profiles at 10 years for MsDiff


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 77 of 112
WP4 REPORT – Modelling of Chloride Ingress

Importance factors

Table 7.7 : Importance factor in time evolution for MsDiff

Importance factors
T (years)
Csurf D
4 62.7 37.3
5 72.8 27.2
6 81.5 18.5
7 87.8 12.2
8 92.0 8.0
9 94.6 5.4
10 96.3 3.7

100
90
80
percentage (%)

70
60 Csurf
50
40 D
30
20
10
0
0 5 10
T (years)

Figure 7.11 Importance factors in time evolution for MsDiff

As the two previous cases, the importance factors are time-dependent and the variation of the
chloride surface concentration is more important than the one of the chloride coefficient.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 78 of 112
WP4 REPORT – Modelling of Chloride Ingress

Curve of the limit state function

2,20E-12

2,00E-12 None passivation


zone
1,80E-12
D (m2/s)

4 years

1,60E-12 10 years
Passivation
1,40E-12 zone

1,20E-12

1,00E-12
15 20 25 30 35 40 45

Csurf (g/l)

Figure 7.12 Curve of the limit state function for MsDiff

7.6 Conclusions
The sensitivity analysis with probabilistic methods has been done on two input data: the
surface chloride concentration and the chloride diffusion coefficient. This study shows that
the importance factors are depending on time. This remark is valid for the three tested models
in spite of their very different conception.
In the considered case, the influence of chloride surface concentration is more important than
the one of diffusion coefficient for the three tested models. So that, we spend more efforts to
increase the precision of the chloride surface concentration measure in order to increase the
precision of the prediction. The chloride surface concentration is function of the environment
but also of the interactions between cement paste and chloride. So the knowledge of the
binding isotherm is a key of the prediction of chloride ingress as important as the diffusion
coefficient.

The above conclusions are based on the sensitivity analysis done on an exposure time up to
ten years. The conclusions are depending on the exposure time. For longer exposure times,
such as 100 years, the conclusions may be different. This is tested in the next chapter.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 79 of 112
WP4 REPORT – Modelling of Chloride Ingress

8 LONG-TERM SENSITIVITY OF ERFC-MODELS

A second sensitivity analysis has been done in sucha way that it covers also much longer
exposure times than 0-10 years. This analysis has been done by Tang Luping at SP.

8.1 Mathematical expressions


Since the initial chloride concentration Ci in an ERFC model only change the vertical
position of a chloride distribution profile, but not influence the shape of the profile, it is
reasonable to assume that Ci ≈ 0, so as to simplify the mathematical expressions. Thus all the
models based on ERFC can be expressed in a general form as follows:

C = C s ⋅ erfc( z ) [8:1]

where C could be further expressed as (C - Ci) and Cs as (Cs - Ci) in the application of the
models. In equation [8:1],

x
z= [8:2]
n′
⎛ t′ ⎞
2 kD0 ⎜ 0 ⎟ ⋅ t
⎝t ⎠

or more correctly,

x
z= [8:3]
kD0 ⎛ t 0′ ⎞ ⎡⎛ t ex ′ ⎞ ⎤
n 1− n 1− n
′ ⎞ ⎛ t ex
2 ⎜ ⎟ ⎢⎜1 + ⎟ − ⎜ ⎟ ⎥ ⋅ t
1 − n ⎝ t ⎠ ⎣⎢⎝ t ⎠ ⎝ t ⎠ ⎦⎥

For model A1c, it is the combination of equations [8:1] and [8:2], while for models A1b and
A1d, let k = 1 in equation [8:2], and for model A1a, let k = 1 and n = n’ = 0 in either equation
[8:2] or [8:3].

Even though in model A1d the surface concentration Cs is a function of time t, it will be seen
later that this will only influence the magnitude of Cs, but not the sensitivity of Cs for
prediction of C.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 80 of 112
WP4 REPORT – Modelling of Chloride Ingress

8.2 Sensitivity of various parameters for prediction of


chloride concentration
8.2.1 Surface concentration Cs
The sensitivity coefficient of parameter Cs is as follows:

∂C
= erfc( z ) [8:4]
∂C s

Therefore,

ΔC C s
⋅ =1 [8:5]
ΔC s C

This implies that any relative change in surface concentration Cs will result in an equal
relative change in concentration C. From equation [8:5] it can be seen that the value of
ΔC C s
⋅ has no dependence on any other parameter, such as t. Therefore, no matter if Cs is
ΔC s C
time-dependent or not, its sensitivity is the same.

8.2.2 Diffusion coefficient D0


The sensitivity coefficient of parameter D0 can be derived as:

∂C C z 2
= 0 ⋅ ⋅ e −z [8:6]
∂D0 π D0

resulting in the following equation:


2
ΔC D0 1 Cs 2 1 z ⋅ e−z [8:7]
⋅ = ⋅ ⋅ z ⋅ e−z = ⋅
ΔD0 C π C π C
Cs

Apparently, the sensitivity of parameter D0 is related to the ratio of C to Cs (or C/Cs) and the
value of z that contains all the variations except for Cs. It can be seen from equation [8:7] that
ΔC D0 ΔC D0
⋅ is inversely proportional to C/Cs. The quantitative relationships of ⋅ and z
ΔD0 C ΔD0 C
ΔC D0
are shown in Figure 8.1. The maximal influence of z on ⋅ appears at z ≈ 0.7. It is
ΔD0 C
relatively less sensitive if the C/Cs is larger than 0.1 and z value is larger than 1, the latter
implies a thicker cover x or longer exposure time t.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 81 of 112
WP4 REPORT – Modelling of Chloride Ingress

25
C/Cs = 0.01
C/Cs = 0.02
20 C/Cs = 0.03
C/Cs = 0.05
ΔC D0 15

C/Cs = 0.1

ΔD0 C C/Cs = 0.2


10 C/Cs = 0.5

0
0 0.5 1 1.5 2 2.5 3

Figure 8.1. Sensitivity of parameter D0 for the prediction of chloride concentration C.

8.2.3 Parameter k
The sensitivity coefficient of parameter k is similar to that of parameter D0, that is,

∂C C 0 z − z 2
= ⋅ ⋅e [8:8]
∂k π k

resulting in an equation equal to equation [8:7]:


2
ΔC k ΔC D 0 1 z ⋅ e−z [8:7’]
⋅ = ⋅ = ⋅
Δk C ΔD0 C π C
Cs

Therefore, the relations in Figure 8.1 are also valid for parameter k.

8.2.4 Parameter n’ in equation [8:2]


The sensitivity coefficient of parameter n’ in equation [8:2] is

∂C C 0 2 ⎛ t′ ⎞
= ⋅ z ⋅ e − z ln⎜ 0 ⎟ [8:9]
∂n′ π ⎝t ⎠

which results in

ΔC n ′ n ′ z ⋅ e − z ⎛ t 0′ ⎞
2

⋅ = ⋅ ln⎜ ⎟ [8:10]
Δn ′ C π C ⎝t ⎠
Cs
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 82 of 112
WP4 REPORT – Modelling of Chloride Ingress

Obviously, besides the influences of C/Cs and z values, which is similar to those discussed in
ΔC n′
section 2.2, ⋅ is directly proportional to the value of n’ and logarithmically
Δn ′ C
proportional to the ratio of t0/t. In the case of C/Cs = 0.1 and z = 1, the relationships of
ΔC n′
⋅ and n’ can be shown in Figure 8.2.
Δn′ C

0 t' 0/t = 0.001

0.002

-5 0.005

0.01

0.02
-10
0.05

0.1
-15
0 0.2 0.4 0.6 0.8 1
n' in equation [8:2]

Figure 8.2. Sensitivity of parameter n in equation [8:2] for the prediction of chloride
concentration C: Influence of value n when C/Cs = 0.1 and z = 1.

8.2.5 Parameter n in equation [8:3]


The sensitivity coefficient of parameter n in equation [8:3] is

⎡ ⎛ t′ ⎞ ⎛ t ex ′ ⎞
1− n
⎛ t′ ⎞ ⎛ t′ ⎞
1− n

⎢ ln⎜1 + ex ⎟ ⋅ ⎜1 + ⎟ − ln⎜ ex ⎟ ⋅ ⎜ ex ⎟ ⎥
∂C C 0 ⎛ t′ ⎞ 1 t t ⎠ ⎝ t ⎠ ⎝ t ⎠
⋅ z ⋅ e − z ⎢ln⎜ 0 ⎟ + − ⎝ ⎠ ⎝ ⎥
2
= [8:11]
∂n π ⎢ ⎝ t ⎠ 1− n ⎛ t ex ′ ⎞
1− n
′ ⎞
⎛ t ex
1− n ⎥
⎢ ⎜1 + ⎟ − ⎜ ⎟ ⎥
⎢⎣ ⎝ t ⎠ ⎝ t ⎠ ⎥⎦

which results in

⎡ ⎛ t′ ⎞ ⎛ t ex ′ ⎞
1− n
⎛ t′ ⎞ ⎛ t′ ⎞
1− n

⎢ ln⎜1 + ex ⎟ ⋅ ⎜1 + ⎟ − ln⎜ ex ⎟ ⋅ ⎜ ex ⎟ ⎥
⎢ln⎛⎜ t 0′ ⎞⎟ + 1 − ⎝
2
ΔC n n z ⋅ e −z t ⎠ ⎝ t ⎠ ⎝ t ⎠ ⎝ t ⎠ ⎥ [8:12]
⋅ = ⋅
Δn C π C ⎢ ⎝ t ⎠ 1− n ⎛ t ex ′ ⎞
1− n
′ ⎞
⎛ t ex
1− n ⎥
⎢ ⎜1 + ⎟ − ⎜ ⎟ ⎥
Cs
⎣⎢ ⎝ t ⎠ ⎝ t ⎠ ⎦⎥
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 83 of 112
WP4 REPORT – Modelling of Chloride Ingress

ΔC n
In this case the influences of n and t’0/t on ⋅ become complicated, and can be
Δn C
illustrated in Figure 8.3, where C/Cs = 0.1 and z = 1.

4 t' 0/t = 0.001

2 0.002

ΔC n 0 0.005

Δn C -2 0.01

0.02
-4
0.05
-6
0.1
-8
0 0.2 0.4 0.6 0.8 1 t' ex/t = 0.001
n in equation [8:3]

Figure 8.3. Sensitivity of parameter n in equation [8:3] for the prediction of chloride
concentration C: Influence of value n when C/Cs = 0.1 and z = 1.

8.2.6 Parameter t0
The sensitivity coefficient of parameter t0 in both equations [8:2] and [8:3] will be similar,
that is,
2
∂C C 0 ⋅ n z ⋅ e − z
= ⋅ [8:13]
∂t 0′ π t 0′

resulting in

ΔC t 0′
2
n z ⋅ e−z
⋅ = ⋅ [8:14]
Δt 0′ C π C
C0

Clearly, besides the influences of C/Cs and z values, which is similar to those discussed in
ΔC t 0′
section 2.2, ⋅ is directly proportional to the value of n. In the case of C/Cs = 0.1 and z
Δt 0′ C
ΔC t 0′ 10
= 1, ⋅ = ⋅ n = 3.18n .
Δt 0′ C π
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 84 of 112
WP4 REPORT – Modelling of Chloride Ingress

8.3 Effect of different parameters on the sensitivity


In order to compare the effect of different parameters on the sensitivity, we have to make the
parameter z constant, because this parameter is directly related to chloride concentration C.
For instance, in the case of x = 0.05 m and t = 100 years, it is not difficult to keep z = 1
through the adjustment of parameters D0, n’ or n.

Under the conditions of z = 1 and t’0/t = 0.005 (for parameter n only, corresponding to t’0 =
0.5 years and t = 100 years), the effects of different parameters on the sensitivity can be
summarise in Figure 8.4.

It can be seen that, among different parameters, n or n’ is the most sensitive one, especially
when its value is larger than 0.2.

Except for parameter Cs, which is independent of other parameters, the sensitivity of all the
other parameters is dependent on the ratio C/Cs.

Cs
4
Do or k
2 n = 0.2
n' = 0.2
0
n = 0.5
ΔC f i -2 n' = 0.5

Δf i C -4
n = 0.8
n' = 0.8
-6 t0 (n' = 0.2)
t0 (n' = 0.5)
-8
t0 (n' = 0.8)
-10
0.1 0.2 0.3 0.4 0.5 n' as in Eq [8:2]
C /C s n as in Eq [8:3]

Figure 8.4. Effects of different parameters on the sensitivity under the conditions of z = 1
and t’0/t = 0.005 (for parameter n or n’ only).

8.4 Discussions
Under the typical exposure conditions, Cs may be about 5 % of binder. If the threshold
chloride content is 0.5 % of binder (e.g. in splash zone), C/Cs will be about 0.1. In this case,
the sensitivity of parameters D0 and any types of k will be 2 times as large as that of Cs, while
the sensitivity of n, once it is larger than 0.2, will be more than 2 times as large as that of Cs.
If the threshold chloride content is 1.5 % of binder (e.g. in submerged zone), C/Cs will be
about 0.3. Thus the sensitivity of parameters D0 and any types of k will be less than that of Cs,
but the sensitivity of n may still be larger than that of Cs.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 85 of 112
WP4 REPORT – Modelling of Chloride Ingress

8.5 Combined uncertainty of models for prediction of


chloride concentration
According to ISO standard expression of uncertainty, 1993 (E), the combined standard
uncertainty uc is the positive square root of the combined variance u c2 , which is given by

2
⎛ ∂C ⎞
N
u = ∑ ⎜⎜
2
c
⎟⎟ ⋅ u 2f i [8:15]
i =1 ⎝ ∂f i ⎠

where f denotes any type of parameter which may contribute to the uncertainty.

The coefficient of variation COV is defined as

uc
COV = [8:16]
C

8.5.1 Model A1a


In model A1a, two parameters, Cs and D0, will contribute to the uncertainty, that is,

2 2
⎛ ∂C ⎞ ⎛ ∂C ⎞
u 2
A1a = ⎜⎜ ⎟⎟ ⋅ u C2s + ⎜⎜ ⎟⎟ ⋅ u D2 0 [8:17]
⎝ ∂C s ⎠ ⎝ ∂D0 ⎠

2
⎛ ⎞
2 ⎜ ⎟ 2
⎛ uC ⎞ ⎟ ⋅ ⎛⎜ u D0 ⎞⎟
2
1 ⎜ z ⋅ e −z
COVA1a = ⎜⎜ s ⎟⎟ + [8:18]
π⎜ C ⎟ ⎜ D0 ⎟
⎝ Cs ⎠ ⎜ ⎟ ⎝ ⎠
⎝ Cs ⎠

8.5.2 Models A1b and A1d


In models A1b and A1d, four parameters, Cs, D0, n and t0, will contribute to the
uncertainty, that is,
2 2 2 2
⎛ ∂C ⎞ ⎛ ∂C ⎞ ∂C ⎞ ⎛ ∂C ⎞
u 2
A1b,d = ⎜⎜ ⎟⎟ ⋅ u C2s + ⎜⎜ ⎟⎟ ⋅ u D2 0 + ⎛⎜ ⎟ ⋅ u n + ⎜⎜
2
⎟⎟ ⋅ u t20 [8:19]
⎝ ∂C s ⎠ ⎝ ∂D0 ⎠ ⎝ ∂n ⎠ ∂t
⎝ 0⎠

2
⎛ ⎞
2 ⎜ ⎟ ⎧ 2 2 2
⎫⎪
⎛ uC ⎞ ⎟ ⋅ ⎪⎛⎜ u D0 ⎞⎟ + ⎡ln⎛⎜ t 0 ⎛ ut ⎞
2
1 ⎜ z ⋅ e −z ⎞ ⎤
COVA1b,d = ⎜⎜ s ⎟ + ⎟u n ⎥ + n 2 ⎜⎜ 0 ⎟ [8:20]
⎟ π⎜ C ⎟ ⎨⎜ D0 ⎟ ⎢⎣ ⎝ t ⎟ ⎬
⎝ Cs ⎠ ⎜ ⎟ ⎪⎩⎝ ⎠ ⎠ ⎦ ⎝ t0 ⎠ ⎪⎭
⎝ Cs ⎠
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 86 of 112
WP4 REPORT – Modelling of Chloride Ingress

8.5.3 Model A1c


In model A1c, seven parameters, Cs, D0, kt, ke, kc, n and t’0, will contribute to the uncertainty.
The expressions are similar to those for models A1b and A1d, that is,

2 2 2 2 2
⎛ ∂C ⎞ ⎛ ∂C ⎞ ⎛ ∂C ⎞ ∂C ⎞ ⎛ ∂C ⎞
u 2
A1c = ⎜⎜ ⎟⎟ ⋅ u C2s + ⎜⎜ ⎟⎟ ⋅ u D2 0 + ∑ ⎜⎜ ⎟⎟ ⋅ u k2i + + ⎛⎜ ⎟ ⋅ u n + ⎜⎜
2
⎟⎟ ⋅ u t20 [8:21]
⎝ ∂C s ⎠ ⎝ ∂D0 ⎠ ∂k
⎝ i⎠ ⎝ ∂n ⎠ ∂t
⎝ 0⎠

2
⎛ ⎞
2 ⎜ ⎟ ⎧ 2 2 2 2
⎫⎪
⎡ u Cs ⎤ ⎟ ⋅ ⎪⎛⎜ u D0 ⎞⎟ + ⎛⎜ u ki ⎞ ⎡ ⎛ t0 ⎛ ut ⎞
2
1 ⎜ z ⋅ e −z ⎞ ⎤
⎟ ⎨⎜ D0 ⎟ ∑ ⎜ k i
COVA1c = ⎢ ⎥ + ⎟ + ⎢ln⎜ ⎟u n ⎥ + n 2 ⎜⎜ 0 ⎟ ⎬ [8:22]
π⎜ C ⎟ ⎟
⎣ Cs ⎦ ⎜ ⎟ ⎪⎩⎝ ⎠ ⎝ ⎠ ⎣ ⎝t ⎠ ⎦ ⎝ t0 ⎠ ⎪⎭
⎝ Cs ⎠

where ki represents kt, ke and kc, respectively.

Apparently, the more parameters in a model, the more sources of error. However, the actual
COV for each model is dependent on the actual magnitude of each individual uncertainty.
Besides, the most important factor for a prediction model is the degree of agreement with the
actual chloride profiles.

8.6 Conclusions
ΔC C s
Parameter Cs. has constant sensitivity in concentration prediction, that is, ⋅ = 1.
ΔC s C
The sensitivity of parameter D0 or k (as a multiplier to D0) is related to the ratio C/Cs and the
value of z that contains all the variations except for Cs. If z = 1, when C/Cs is larger than 0.2,
the sensitivity of D0 or k in concentration prediction becomes less than parameter Cs,
Otherwise it becomes larger than Cs. On the other hand, when the value of z is very small or
very large (not close to 0.7), the prediction of chloride concentration also becomes less
sensitive to parameter D0 or k.

Among all the parameters, the age factor n or n’ is the most sensitive one, especially when its
value is larger than 0.2.

Parameter t0 has relatively less sensitivity in the prediction of chloride concentration,


especially if the age factor n or n’ is less than 0.5.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 87 of 112
WP4 REPORT – Modelling of Chloride Ingress

9 BENCHMARKING OF MODELS

9.1 Objectives and over-view of work performed


Task 4.3 Bench Marking, by partners IETcc (task leader), SP, LTH and EdF, concerned
studying the models’ abilities to reproduce reality, with laboratory test results as input and in-
field performance data as comparison.

This was done by selecting some field data, which had the concrete composition and
properties and the environmental conditions well documented, and ask model developers to
make predictions with their models on these field cases. The prediction results were then to
be analyzed by using a probabilistic approach.

The complete background data and prediction results of task 4.3 are presented in a final
working report by the task leader IETcc: “BENCHMARKING OF MODELS”, dated October
25, 2005. This report has been complimented with an “AMENDMENT”, dated January 2006,
by the work package leader LTH, adding and correcting some data and analysis.

The work performed in task 4.3 was the following:


1. Establishment of criteria for the benchmarking
2. Selection of profiles from the data base
3. Selection of the model producers more reputed in the world
4. Submission of the documentation to the modellers
5. Collection of results and application of benchmarking criteria
6. Discussion on results obtained and suggestions for further actions

9.2 Establishment of criteria for benchmarking


Prior to the work, there were established criteria for making the comparison between
predicted profiles and real ones, in collaboration between IETcc and EDF.

The method proposed for making the benchmarking was selected on the basis of the
comparison of the bias of the predicted profiles of each model with respect to the real one,
based in the areas between both profiles as is explained in the next section. It has the
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 88 of 112
WP4 REPORT – Modelling of Chloride Ingress

advantage to quantify in some manner the deviation and therefore to make the models
comparable.

9.2.1 Comparing different models when the measured data is known


To verify the performance of the different mathematical approaches made by different
researchers, a methodology based in the area between the curves was developed. The
prediction result from each model can be compared with the measured curve by comparing
the value of the area between the predicted curve and the measured curve in a depth range
from X1 to X2 as shown in Figure 9.1.

The values used to compare are the Areas with and without sign called:

S1 = ⏐A1⏐+⏐A2⏐ S2 = A1 + A2

c Measured
Predicted
c2

A1
c1
A2

X
x1 x2

Figure 9.1 Predicted and measured profiles of chloride concentration

where :
c1 : measured curve (total chlorides)
c2 : model curve (total chlorides)
A1 : area between the measured curve and model curve which measured data is lower
(negative sign) [%·mm]
A2 : area between the measured curve and model curve which measured data is higher
(positive sign) [%·mm]
x1, x2 : validation depth range
C, X : concentration and depth axis respectively

Further information:
The c1 and c2 curves are cubic spline interpolation of measured and model data
points. If the measured data or predicted profile do not cover the whole interval
between x1 and x2 an extrapolation was made.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 89 of 112
WP4 REPORT – Modelling of Chloride Ingress

S1 gives the information about how near the model is from the measured data.
S2 gives the information about how much higher or lower the predicted curve is,
compared with the measured data.

S1 > 0
S2 < S1 always
S2 < 0 if the measured profile is lower than the predicted
S2 = -S1 if the whole measured profile is lower than the predicted
S2 = S1 if the whole measured profile is higher than the predicted

The validation range adopted was:


• x1 = 10mm ( this value was adopted to not take into account the skin effect)
• x2 = 50mm (except in the situation when the depth of the measured data is smaller).

It is obvious from some cases that this choice of validation range is questionable where the
depth of penetration is small. A significant part of the profile falls then between 0 and x1 and
is excluded from the comparison.

9.2.2 Comparing different models when the measured data is unknown


When the measured data is unknown, for example for predicted chloride profiles at 100 years,
the table with the modellers’ data and the graphics are given, but no comparison is made.

9.3 Selection of profiles and documentation prepared for the


benchmarking
The purpose is to know the deviation of the predicted data from the real data, represented by
a second profile, of the same concrete at a second age. A prediction to 100 years was
introduced in order to evidence the sensitivity of the models to take into account the age
effect on diffusivity.

As the task tries to study the ability of the different available models to predict further
evolution of chloride profiles, the prediction can be made based on results (a profile) from
one age, accompanied or not with short term results in the laboratory, of the same concrete.

The tasks performed were the following:

1) Selection of a set of chloride profiles from the database of ChlorTest following the criteria
given by the leaders of the project (SP) and of WP 4 (LTH). The selection should have been
made by WP 3.
• A set of 400 profiles was sent to the leader of task 4.3 at the beginning of January 2005.
• It was recognized that most of the profiles did not follow the criteria fixed previously in
the meeting of November 2004.
• Only profiles provided by SP seem to meet all the criteria
• It seemed to leaders of task 4.3 that a wider variety of profiles is necessary to assure
enough objectivity to the benchmarking exercise.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 90 of 112
WP4 REPORT – Modelling of Chloride Ingress

Due to the enormous amount of profiles sent to task 4.3 (400) extra work has been to select
some of them to fulfil the criteria given on November 2004.

Therefore, IETcc prepared the following set of profiles:

1. Those of SP because they meet the criteria fixed. Two profiles, each at a different
age, will be submitted as proposed on November 2004. The prediction exercise will
be from the first age profile to the second and to 100 years.
2. A second set of profiles was selected not meeting all the criteria. Only one profile at
one age will be sent to the modellers requesting prediction to the second age (profile
not sent) and 100 years.
3. Data from old real structures will constitute a third set of profiles to be sent to
modellers to predict the 100 year profiles.

2) A brief guide for modellers was prepared together with the profiles. The guide contained:

• Summary of the project


• Purpose of benchmarking exercise (task 4.3)
• The three set of profiles selected with indications of needed predictions and form for
submitting results
• Form for responding on positive or not collaboration with task 4.3
• A questionnaire to specify by the modeller the characteristics of the models
• Excel files with the given profiles with space for the predicted ones.

In Annex A of the final task 4.3 working report the profiles submitted for the benchmarking
are enclosed. Three series of data were selected with indications of needed predictions. The
results of the calculated profiles at the ages requested from each modeller were asked to be
given back in the same Excel files. The three series correspond to 22 different cases.

The conditions of the profiles were:

Series 1: One early of two chloride profiles from two consecutive ages in the same concrete is
given. Results of short term accelerated tests are also included. The modeller is requested:
• to predict the second age profile
• to predict the profile at 100 years

Series 2: One of two chloride profiles from two consecutive ages in the same concrete is
given. No result of short-term accelerated test is available. The modeller is
requested:
• to predict the second age profile
• to predict the profile at 100 years

Series 3: One profile from the old structure is given. The modeller is requested:
• to predict the profile at 100 years

The modellers were encouraged to model all the cases if possible. The modellers can also
model a limited number of cases if their time is limited. In this circumstance, it was suggested
to the modellers to follow the order of the case number.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 91 of 112
WP4 REPORT – Modelling of Chloride Ingress

• For series 1, the exercise must be performed with the 4 first profiles supplied. The rest
of profiles included in series 1 are voluntary
• In series 2 and 3 the exercise is requested in the case of the first profile. The rest of
profiles supplied are voluntary.

In Series 1 the concrete chloride ingress properties were given as a measured chloride migration
coefficient. That kind of information was not available in Series 2 and 3. In series 1 the given
profiles were after a very short time of exposure, less than 10 % of the exposure time for the
required prediction. In Series 2 and 3, however, the profiles were given after some 40 % of the final
exposure time.

The different cases can be summarized in the following way:

Series 1 Exposure time [years]


Case Concrete Environment Given profile Measured/Predicted Comment
C1 CEM I Submerged limited
C2 Splash 0.8 10.3 thickness
C3 w/c 0.4 Atmospheric
C4 CEMI/0.4 Road VF 0.4 4.6 small diff. 0.4-4.6 y
C5 CEM I 5 %Si 10.2
Submerged 1.0
C6 w/b 0.35 20 %FA 10.1
C7 CEMI/0.4 Road H 0.4 4.6 small diff. 0.4-4.6 y
C8 CEM I 5 %Si Submerged
0.7 10.2
C9 w/b 0.35 10 %FA Splash
C10 CEM I Atmospheric
0.5 3.0
C11 w/c 0.5 Tidal
C12 CEM I Road VF
5 %Si 0.4 4.6 small diff. 0.4-4.6 y
C13 w/b 0.4 Road H
VF= vertical surface
H = horisontal surface

Series 2 Exposure time [years]


Case Concrete Environment Given profile Measured/Predicted Comment
C1 OPC CEM I 7 12
Tidal
C2 SRPC w/c 0.5 6 14 small diff. 6-14 y
C3 CEM II/B w/c 0.57 Splash 8 18 small diff. 8-18 y
C4 Submerged
CEMIV/B w/c 0.4 none 1.5
C5 Tidal

Series 3 Exposure time [years]


Case Concrete Environ. Over sea [m] Given profile Measured/Predicted Comment
C1 Tidal 1
C2 Atmos- 9
CEMIII/B 42 none/100
C3
14 no real difference
C4 w/c 0.45 pheric
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 92 of 112
WP4 REPORT – Modelling of Chloride Ingress

9.4 Selection of models


The selection was made attending two criteria:

1. Models developed by reputed individuals which have published in the scientific


literature.

2. Models that are known and currently used, like ”error function” or ”square root”
models.

The list of model developers addressed is the following:

• Ilie Petre-Lazars – EDF- France


• Tetsuya ISHIDA - University of Tokyo - Japan
• Tang Luping - SP-Sweden
• Lars-Olof Nilsson - University of Lund - Sweden
• Myriam Carcasses- INSA, Toulouse- France
• Sander Meijers- Intron - The Netherlands
• Kazuo Yamada - Taiheiyo Cement – Japan
• Jens M. Frederiksen – Birch & Krogboe - Denmark
• Jacques Marchand - Laval University-Canada
• Evan Bentz - University of Toronto - Canada
• Dale Bentz- NIST- USA
• Steinar Helland - Skanska- Norway
• Heidi Ungricht- Switzerland
• Eugen Brühwiler- University of Lausanne - Switzerland
• Joost Gulikers- the Netherlands
• Nick Buenfeld- Imperial College- UK
• Oladis de Rincon - University of Zulia- Venezuela
• Carmen Andrade- IETCC - Spain

9.5 Responses obtained


Not all the modellers could participate, for different reasons. The available time was very
short and just before the summer vacation 2005. Some models require two profiles as input
data and that was not available.

The final number of models tested has been 16, which is thought representative enough for
the purpose of the project. In total there are 16 models from 9 different modellers. The
“model 4” is the traditional Error Function.

Table 9.1 shows the profiles analyzed in this work with the series that have been made.

“1” represents that the profile in the second age was made by the modeller
“2” represents that the profile for 100 years was made by the modeller
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 93 of 112
WP4 REPORT – Modelling of Chloride Ingress

Table 9.1 – Models received from the modellers


Series 1 Series 2 Series3
C1 C2 C3
C4 C5 C6 C7 C8 C9 C10 C11 C12 C13 C1 C2 C3 C4 C5 C1 C2 C3 C4
Mod1 1,2 1,2 1,2
1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1 1 1
Mod2 1,2 1,2 1,2 1,2 1,2 1,2 1,2
Mod3 1,2 1,2 1,2
1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2
Mod4 1,2 1,2 1,2
1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1 1 1
Mod5 1,2 1,2 1,2
1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1 1 1
Mod6 1,2 1,2 1,2
1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1 1 1
Mod7 1,2 1,2 1,2
1,2 1,2 1,2 1,2
Mod8 1,2 1,2 1,2
1,2 1,2 1,2 1,2
Mod9 1,2 1,2 1,2
1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1 1 1
Mod10 1,2 1,2 1,2
1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1 1 1
Mod11 1,2 1,2 1,2 1,2
Mod12 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1,2 1,2 1,2 1,2 1 1 1 1
Mod13 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1 1 1
Mod14 1 1 1 1
Mod15 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1,2 1 1 1 1
Mod16 1,2 1,2 2 1,2 1,2 1,2 1,2 2 1 2

Table 9.2 shows the summary of the models used in this task. The complete answered
questionnaires are found in the Annex B of the final working report.

Table 9.2 – Summary of the questionnaires answered by the modellers.


MODEL
1 2 3 5 6 7 8 9 10 11 12 13 14 15 16
BOTH

BOTH

BOTH

Based on Theoretical Principles [TP]


EC

EC

EC

EC

EC

EC
TP

TP

TP

TP

TP

TP

or on Empirical Calibration[EC]
YES

YES

YES

YES

YES

YES

YES

YES

YES

YES

YES

YES

YES

YES

NO

Based on Fick´s second law


YES

YES

YES

YES

YES

YES

YES

YES

YES
Yes

Yes
NO

NO

NO

NO

Temporal dependence of Cs
YES

YES

YES

YES

YES

YES

YES

YES

YES
NO

NO

NO

NO

NO

NO

Need to introduce a D
BOTH

Type of D : steady-state [SS] or snon-


NSS

NSS

NSS

NSS

NSS
SS

SS

SS

teady-state [NSS]
YES

YES

YES

YES

YES

YES

YES

YES
NO

NO

NO

NO

NO

Chloride binding
YES

YES

YES

YES

YES

YES

YES

YES

YES

YES

YES

YES
NO

NO

NO

Temporal dependence of D

Assume some value when it's not


YES

YES

YES

YES

YES

YES

YES

YES

YES
NO

NO

NO

NO
NO

available

It is noteworthy that 11 of the 16 models are using a surface chloride concentration that
changes with time.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 94 of 112
WP4 REPORT – Modelling of Chloride Ingress

9.6 Comparison of results


The result for the series 1 is presented with the schema below for each Case:
• The Table with the predicted and the measured profile at the second age,
• The Graphic with the predicted and real profile for the Second age,
• The Histogram of the S1 value of the modellers for the Second age,
• The Histogram of the S2 value of the modellers for the Second age,
• Comments about the results of the Second age,
• The Table with the predicted and real profile for the 100 year predictions,
• The Graphic with the predicted and real profile for the 100 year predictions,
• Comments about the results of the 100 year predictions.

All results are documented in the final working report. Some examples are given in the next section. An
analysis is made in the following section.

9.7 Some examples of results

Series 1 – Case 01 – 10.3 years of exposure


Series 1, Case 01, is a CEM1 concrete with a w/c of 0.40, constantly submerged into sea
water. The given profile after an exposure time of 0.78 years is shown in Figure 9.2.
7

6 Series 1 Case 1
Cl/C [wt% of binder]

4
3

2
1

0
0 10 20 30 40 50
Depth (mm)

Figure 9.2.- Series 1- given profile after an exposure of 0.78 years

The given profile has a surface chloride content Cs around 2.5-3 % of binder.

The predicted profiles after an exposure time of 10.3 years are shown in Figure 9.3 together
with the measured profile (“Mod 0”). The differences between the predicted profiles and the
measured one are shown in Figures 9.4 and 9.5, expressed as S1 and S2 values.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 95 of 112
WP4 REPORT – Modelling of Chloride Ingress

Chlortest WP4.3 - Series 1 - Case 1- T=10.3

7,001

6,001

5,001
Chloride Concentration

4,001

3,001

2,001

1,001

0,001
0,00 10,00 20,00 30,00 40,00 50,00
Depth

Mod 0 Mod 1 Mod 2 Mod 3 Mod 4 Mod 5 Mod 6 Mod 7


Mod 8 Mod 9 Mod 10 Mod 12 Mod 13 Mod 15 Mod 16

Figure 9.3.- Series 1- profile 1. The profile showing the maximum (Mod 0 ) is the measured profile

Chlortest WP4.3 - Series 1 - Case 1-


T=10.3
Mod 1
Mod 2
80,00
Mod 3
70,00
Mod 4
60,00 Mod 5
50,00 Mod 6
S1

40,00 Mod 7
30,00 Mod 8

20,00 Mod 9
Mod 10
10,00
Mod 12
0,00
Mod 13
Mod 15
Models
Mod 16

Figure 9.4- Series 1- profile 1.Value of area S1 between real and predicted profiles.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 96 of 112
WP4 REPORT – Modelling of Chloride Ingress

Chlortest WP4.3 - Series 1 - Case 1- T=10.3


Mod 1
100,00
Mod 2
80,00 Mod 3
60,00 Mod 4
Mod 5
40,00
Mod 6
S2

20,00 Mod 7

0,00 Mod 8
Mod 9
-20,00
Mod 10
-40,00 Mod 12
Mod 13
-60,00
Mod 15
Models
Mod 16

Figure 9.5- Series 1- profile 1.Area S2 between real and predicted profiles

Comments on Series 1, Case 1, 10.3 years


The first remark is that the profile after 10.3 years does not represent a semi-infinite case.
The profiles have been taken from specimens with a thickness of only 100 mm and
penetration from two sides. For an exposure time of 10.3 years, the penetration from each
surface is more than 50 mm. Consequently, the measured chloride profile is not fully relevant
for comparing with the predicted profiles, since all predictions are made for semi-infinite
cases. Having this in mind, the results could be analyzed with some limitations.

There are large differences between the predicted profiles. Roughly, there are some models
(1, 5, 9, 10 and 13) that have correctly predicted the change with time of Cs and therefore
they fit the measured profile much better. Others, not having picked the new Cs, in reality
used a constant Cs-value, predict profiles very far from the measured profile in the initial part
of the profile near the surface.

It is also interesting to notice that there are large differences between the predicted profiles
within the two groups using a time-dependent or constant Cs-value, respectively. Even
though the input data included a given, “early” profile and a D, the D used in the predictions
are very different.

However, it is interesting to realize that there are not large differences in the ”front part” of
the predicted profiles of those which picked more or less well the new Cs with those
considering a constant Cs. That is, at 5 cm depth the predicted profiles do not differ too
much. This is of course due to the low D in all cases.

As summary, in spite of the difficulty in the prediction, there are profiles which reach very
good predictions in this case. These are mainly MOD 5 and MOD 9.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 97 of 112
WP4 REPORT – Modelling of Chloride Ingress

Comments on Series 1, Case 1, 100 years


Here the prediction is up to 100 years, see Figure 9.5. Now the prediction at 5 cm of course
differs much more between the models that have considered an increase of Cs with time
compared to those which did not consider this. Therefore, the Cs-change has a strong
influence in the long term prediction. There is also a significant difference in the D:s used.

Chlortest WP4.3 - Series 1 - Case 1- T=100

10,00

8,00
Chloride Concentra

6,00

4,00

2,00

0,00
0,00 10,00 20,00 30,00 40,00 50,00
Depth

Mod 1 Mod 2 Mod 3 Mod 4 Mod 5 Mod 6 Mod 7


Mod 8 Mod 9 Mod 10 Mod 12 Mod 13 Mod 15 Mod 16

Figure 9.6- Series 1- profile 1. 100 years exposure

Comments on Series 1, Case 4, 4.6 years


Series 1, Case 4, is a similar concrete as in case 1 but now exposed in a road environment.
The given profile is measured after 0.4 years, i.e. during the first year, see Figure 9.7.
Actually, the profile was taken just after the first winter, before the rains in the summer and
fall, when chloride is leached out.

The predicted profiles after 4.6 years are shown in Figure 9.8 together with the measured
profile. All predicted profiles are overestimating the chloride contents after 4.6 years! The
measured profile after 4.6 years has actually lower chloride contents in the outer part of the
concrete that the measured profile after 0.4 years!
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 98 of 112
WP4 REPORT – Modelling of Chloride Ingress

6 Series 1 Case 4
Cl/C [wt% of binder] 5

4
3

2
1
0
0 10 20 30 40 50
Depth (mm)

Figure 9.7- Series 1- case 4. given profile after 0.4 years of exposure

Chlortest WP4.3 - Series 1 - Case 4- T=4.6

8,001

7,001

6,001
Chloride Concentration

5,001

4,001

3,001

2,001

1,001

0,001
0,00 10,00 20,00 30,00 40,00 50,00
Depth

Mod 0 Mod 1 Mod 3 Mod 4 Mod 5 Mod 6 Mod 7


Mod 8 Mod 9 Mod 10 Mod 11 Mod 12 Mod 13

Figure 9.8- Series 1- profile 4. 4.6 years exposure

Here the measured profile presents very low values and therefore the prediction is more
accurate by those models that consider a constant Cs-value equal to the given one. Those
assuming an increase of Cs with time have strongly failed.

The models having achieved a prediction closest to the measured profile are: MOD: 3, 8 and
10.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 99 of 112
WP4 REPORT – Modelling of Chloride Ingress

9.8 Analysis of all predictions

9.8.1 Series 1
The other cases in series 1 gave similar results. The S1-values for all the cases in series 1 are
shown in Table 9.3 and the S2-values in Table 9.4. The tables also give the average S1- and
S2-value, and the standard deviation for each model and each case.

Table 9.3 Over-view of all predictions for series 1 compared to the measured profiles, expressed
as S1-values.
Series 1 C1 C2 C3
C4 C5 C6 C7 C8 C9 C10 C11 C12 C13
Mod S1 S1 S1
S1 S1 S1 S1 S1 S1 S1 S1 S1 S1 Average StDev
1 21.2 50.3 9.1
12.1 5.0 16.7 12.8 30.7 5.6 7.1 17.9 6.9 3.8 15.3 13.0
2 34.2 63.1 11.0 8.4 25.4 21.1 8.4 24.5 19.6
3 75.2 69.7 11.6
13.4 10.2 21.0 15.1 21.4 7.0 21.8 25.0 10.1 12.9 24.2 22.1
4 35.0 65.0 9.6
15.8 6.6 24.5 14.9 20.4 8.0 6.1 16.7 16.5 20.4 20.0 15.7
5 15.8 12.4 1.6
6.1 4.3 6.8 7.8 6.5 1.9 5.8 9.9 2.5 1.5 6.4 4.3
6 66.8 46.3 49.3
63.8 9.9 8.6 59.2 18.8 27.7 5.0 7.7 53.0 59.3 36.6 24.0
7 52.4 25.3 76.6
89.4 7.9 10.2 43.6 34.6
8 36.3 65.5 13.5
30.8 6.2 23.4 29.3 20.9
9 13.7 72.3 4.2
13.0 7.9 18.6 12.7 18.4 3.3 6.4 18.0 8.8 8.3 15.8 17.7
10 36.2 74.4 2.8
6.5 3.0 10.5 3.3 9.5 8.7 6.0 22.0 2.5 5.5 14.7 20.3
11 22.5 18.9 24.1 30.9 24.1 5.0
12 50.5 76.4 3.1 6.8 9.4 9.2 7.4 20.6 2.3 7.5 17.7 3.9 7.8 17.1 21.9
13 47.3 44.4 19.3 44.6 4.7 9.0 46.3 19.6 4.6 4.6 29.3 10.4 12.8 22.8 17.3
14
15 68.2 84.4 2.7 12.4 10.7 19.1 2.3 10.7 19.2 25.5 29.7
16 49.4 69.1 6.3 21.9 16.6 12.5 29.3 24.6
Average 43.0 58.5 16.5 27.1 7.3 15.5 19.8 18.6 7.3 8.1 17.8 13.9 16.3
StDev 19.1 20.4 22.0 26.3 2.6 6.8 18.2 6.0 7.2 5.1 6.3 15.3 17.4

Table 9.4 Over-view of all predictions for series 1 compared to the measured profiles, expressed
as S2-values.
Series 1 C1 C2 C3
C4 C5 C6 C7 C8 C9 C10 C11 C12 C13
Mod S2 S2 S2
S2 S2 S2 S2 S2 S2 S2 S2 S2 S2 Average StDev
1 19.5 50.3 -9.1
-12.1 -5.0 16.7 -12.8 30.7 -5.6 -7.1 16.0 6.9 3.8 7.1 18.9
2 34.2 63.1 -11.0 3.8 -25.1 -4.2 -8.3 7.5 30.6
3 75.2 69.7 -11.6
-13.4 8.1 21.0 -15.1 7.9 -6.6 -21.8 -12.0 -9.7 -12.9 6.1 31.7
4 35.0 65.0 -9.6
-15.8 1.2 -24.5 -14.9 -3.8 -8.0 0.7 -0.7 -16.5 -20.4 -0.9 24.8
5 4.3 -1.1 1.3
-6.1 -0.6 -5.8 -7.8 -3.4 1.3 -0.7 0.6 2.5 -0.6 -1.2 3.6
6 66.8 46.3 -49.3
-63.8 7.9 4.1 -59.2 17.5 -27.7 -5.0 -2.1 -53.0 -59.3 -13.6 42.5
7 52.4 25.3 -76.6
-89.4 5.8 -6.8 -14.9 56.6
8 36.3 65.5 -13.5
-30.8 2.1 -23.2 6.1 37.6
9 4.1 72.3 -3.3
-13.0 7.9 -18.6 -12.7 -15.7 -3.3 1.4 2.5 -8.8 -8.3 0.4 23.1
10 36.1 74.4 -0.9
-6.2 1.6 10.4 -2.8 2.9 -8.7 2.9 20.5 -1.4 -5.1 9.5 23.0
11 -22.5 -18.9 -24.1 -30.9 -24.1 5.0
12 50.5 76.4 -2.7 -6.8 9.4 7.9 -7.4 20.4 2.2 7.4 13.1 -3.6 -7.6 12.2 24.9
13 -47.3 44.4 -19.3 -44.6 4.5 8.9 -46.3 19.6 4.6 -3.6 29.3 -6.6 -9.7 -5.1 28.7
14
15 68.2 84.4 2.7 12.4 10.7 19.1 2.3 10.7 17.7 25.4 29.7
16 49.4 69.1 1.9 -16.6 4.5 5.1 18.9 32.9
Average 34.6 57.5 -15.6 -27.0 4.4 -2.9 -19.8 8.0 -5.3 -1.5 8.2 -11.4 -15.1
StDev 32.1 22.9 22.6 26.3 4.6 16.3 18.2 13.6 8.9 8.9 12.1 17.1 18.4

The colours in the tables give some additional information, compared to the final task 4.3
report:
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 100 of 112
WP4 REPORT – Modelling of Chloride Ingress

data missing; now calculated and added


data wrong, now corrected
best comparison

Some of cases (5, 9 and 10) gave excellent comparison between predicted and measured
profiles, even though a few of the models predicted profiles that are far from the measured
ones. Cases 1-2 gave very bad results for most of the models. The extreme increase of Cs
with time, from 0.8 years to 10 years was difficult to predict.

One model is outstanding in this comparison, Model 5. However, this model has been
compared to most of the data of Series 1 and adjusted accordingly. Four more models are in a
second group, Models 1, 9, 10 and 12. They are, however, far from the measured profile in
some of the cases.

This way of comparing models are, however, doubtful. The magnitude of the S1- and S2-
values depends on the magnitude of the chloride contents for each profile. A profile with
high chloride concentrations will give much higher S1-values than a profile with small
chloride contents. Consequently, the different cases will contribute differently to the
comparison, which can be questioned.

9.8.2 Series 2
The comparison between measured and predicted profiles are shown in Tables 9.5 and 9.6.
Some predicted profiles are so unexpected low, that they have not been quantified. This is
because of a mistake in the originally given profiles that were expressed in % by weight of
concrete instead of % of binder. Some of the modellers did not change their predictions after
new profiles were distributed.

The number of predictions was small for cases 4 and 5. These cases are not further dealt with.

Table 9.5 Over-view of all predictions for series 2 compared to the measured profiles, expressed
as S1-values.
Series 2 C1 C2 C3 C4 C5
Mod S1 S1 S1 S1 S1 Average StDev
1 29.4 42.4 94.4 8.7 10.2 37.0 35.0
4 45.3 34.6 8.5 29.4 19.0
5 12.7 11.9 7.9 11.3 16.0 12.0 2.9
6 9.0 9.0 11.8 9.9 1.6
7 13.3 13.3
8 10.3 10.3
9 10.6 11.5 11.3 11.1 0.5
10 18.9 7.9 3.3 10.0 8.0
15 X X 4.5 4.5
16 26.6 26.6
Average 18.7 20.6 20.2 10.0 13.1
StDev 12.6 13.9 32.9 1.8 4.1
first x1=15mm!
first x1=20mm!
X data missing; based on original erroneous data
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 101 of 112
WP4 REPORT – Modelling of Chloride Ingress

Table 9.6 Over-view of all predictions for series 2 compared to the measured profiles, expressed
as S2-values.
Series 2 C1 C2 C3 C4 C5
Mod S2 S2 S2 S2 S2 Average StDev
1 -3.6 -41.5 -94.4 -8.7 0.1 -29.6 39.8
4 45.3 34.6 -8.4 23.8 28.4
5 0.0 -11.8 -7.9 -11.3 16.0 -3.0 11.6
6 1.5 -9.0 -11.8 -6.4 7.0
7 8.2 8.2
8 10.3 10.3
9 3.5 -11.5 -10.9 -6.3 8.5
10 13.3 -3.8 2.9 4.1 8.6
15 X X -2.7 -2.7
16 26.6 26.6
Average 9.8 -2.3 -19.0 -10.0 8.1
StDev 15.4 25.7 33.6 1.8 11.2
first x1=15mm!
first x1=20mm!

Once again, Model 5 gave good predictions, now together with Models 9 and 10. These cases
are ”new” for Model 5, contrary to the ones in Series 1. Other model gave some good
prediction but only dealt with one or two cases.

9.8.3 Series 3
Series 3 had given profiles after 42 years as “calibration” of the models. In spite of this, the
comparison between predicted profiles after 100 years of exposure showed significant
differences.

One example is shown in Figure 9.9, case 1 in Series 3. It is a case with a concrete exposed in
the tidal zone, which should be ”easier” than cases 2-4 which are in the atmospheric zone.

The given profile after 42 years had a Cs-value of some 2.5. The predicted surface chloride
contents vary from some 2.5, over around 4.0 up to 12.0 and above 30! It is also clear from
the profiles that the predictions utilize very different diffusion coefficients even though a
calibration was possible after 42 years.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 102 of 112
WP4 REPORT – Modelling of Chloride Ingress

Figure 9.9- Series 3- case 1 after 100 years exposure, with given profiles after 42 years as
“calibration”.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 103 of 112
WP4 REPORT – Modelling of Chloride Ingress

9.9 Final comments on the benchmarking of models


The most relevant remarks to be made are:
1. Regarding the sensitivity of the models to the effect of the constitutive parameters, it
seems possible to deduce from the study that one of the most influencing one is the Cs
value, both at short and at long term (100 years). This parameter was also identified in
task 4.2. Task 4.3 confirms the importance of correct assumptions on the boundary
conditions for making reliable predictions.
2. Almost all the models have Cs as the input, except a few models which have the free c
as the input and convert free c to Cs with binding factors. The larger influence of Cs
and less influence of D from the study is an indication that the assumption of a
constant Cs in most of models is far away from the reality.
3. Better models to predict the environmental actions are urgently required.
4. With respect to D-values, they are much less influencing than the Cs-values as was
also identified in task 4.2. One aspect has to be mentioned about D values: they were
given to the modellers only in some of the profiles submitted (series 1), which means
that the modellers have taken the D-values they have considered the most convenient
in the rest of series.
5. Where the same surface chloride conditions have been used, very different D-values
have been used in different models, even though an initial profile and a measured D
was given as input data. The models obviously treat chloride diffusion coefficients in
very different ways.
6. Finally with respect to the methodology used to make the benchmarking, the cases
could have been better chosen and the analysis could have been more thorough, but it
has been shown very helpful to compare the measured and predicted profiles.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 104 of 112
WP4 REPORT – Modelling of Chloride Ingress

10 CONCLUSIONS
The Work Package 4 of ChlorTest has three objectives:
• Analysis of models,
• Correlation of lab tests to in-field performance,
• Bench marking of models (regarding EU standardization needs (EN206, EC2)

In reality these objectives may be summarized in one sentence: Make sure/exemplify that test
results can be used (in practice)!

The work package has been divided into three tasks, 4.1 - 4.3, with the objectives “Critical
evaluation of models” (4.1), “Sensitivity analysis of models” (4.2) and “Benchmarking of
models” (4.3).

Conclusions from the critical analysis of models


General conclusions on models based on Fick’s 2nd law are

• The effect of the other ions than chloride in the pore system is completely neglected
in empirical models solving Fick’s 2nd law.
• The meaning of regression parameters in empirical models is not always easy to
understand. They must be very clearly defined.
• The background for a number of the assumptions made in empirical models could be
questioned. This is especially important for the continuous time-dependency of the
diffusion coefficients and the surface chloride contents.
• Empirical models need huge data bases: a Cs for every concrete X, every road Y and
every ”water” Z!

These conclusions give rise to questions on whether empirical models can be used at all for
predictions beyond where data exists. The last conclusion gives rise to questions on whether
empirical models will ever be practical to use for a new structure, made of a new concrete
exposed in a new environment!

The main conclusions on the models based on flux equations are

• Physical models, especially those considering all ions and convection at the same
time, give theoretical predictions as close to our present knowledge as possible.
• Physical models may be used to check the relevance of more simple, empirical
models and may be explain some the assumptions that have to made in simple models.
That may improve the confidence in empirical models.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 105 of 112
WP4 REPORT – Modelling of Chloride Ingress

• Physical models need quantification of a number of material parameters and a number


of environmental parameters, each as a function of several parameters,
• The required surface climatic conditions are extremely complicated to model.
• Because of lack of availability, limited user-friendliness and the huge amount of
required input data most physical models will remain research tools, not meant for
practical applications.

Conclusions on the sensitivity analysis of models


The sensitivity analysis in task 4.2 on how each model appraises environmental differences,
concrete characteristics and corrosion onset was dealt with theoretically, first with a
probabilistic approach on the first 10 years ingress for a selection of models and then in a
general way on the ERFC-models.

The sensitivity analysis with probabilistic methods has been done on two input data: the
surface chloride concentration and the chloride diffusion coefficient. This study shows that
the importance factors are depending on time. This remark is valid for the three tested models
in spite of their very different conception.

In the considered case, the influence of chloride surface concentration is more important than
the one of diffusion coefficient for the three tested models. So that, we spend more efforts to
increase the precision of the chloride surface concentration measure in order to increase the
precision of the prediction. The chloride surface concentration is function of the environment
but also of the interactions between cement paste and chloride. So the knowledge of the
binding isotherm is a key of the prediction of chloride ingress as important as the diffusion
coefficient.

The above conclusions are based on the sensitivity analysis done on an exposure time up to
ten years. The conclusions are depending on the exposure time. For longer exposure times,
such as 100 years, the conclusions may be different.

The long-term sensitivity analysis shows, however, that the influence of Cs is constant, while
the influence of D is dependent on the C/Cs. The most sensitive parameter is the age factor in
some empiric models.

Conclusions on benchmarking models


The bench-marking in task 4.3 aimed at studying the models’ abilities to reproduce reality,
with laboratory test results as input and in-field performance data as comparison.

The most relevant conclusions are:


• Regarding the sensitivity of the models to the variation of the constitutive parameters, it
seems possible to deduce from the study that one of the most influencing parameter is the
Cs value, both at short and at long term (100 years). This parameter was also identified in
task 4.2. Task 4.3 confirms the importance of correct assumptions on the boundary
conditions for making reliable predictions.
• With respect to D-values, they are much less influencing than the Cs-values as was also
identified in task 4.2. One aspect has to be mentioned about D values: they were given to
the modellers only in some of the profiles submitted (series 1), which means that the
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 106 of 112
WP4 REPORT – Modelling of Chloride Ingress

modellers have taken the D-values they have considered the most convenient in the rest of
series.
• Almost all the models have Cs as the input, except a few models which have the free c as
the input and convert free c to Cs with binding factors. The larger influence of Cs and less
influence of D from the study is an indication that the assumption of a constant Cs in most
of models is far away from the reality.
• Where the same surface chloride conditions have been used, very different D-values have
been used in different models, even though an initial profile and a measured D was given
as input data. The models obviously treat chloride diffusion coefficients in very different
ways.
• The benchmarking results show that one model (5), which was previously calibrated with
most of the profiles from Series 1, gives reasonable good prediction, indicating the
importance of calibration with the reliable long-term data.
• Finally with respect to the methodology used to make the benchmarking, the cases could
have been better chosen and the analysis could have been more thorough, but it has been
shown very helpful to compare the measured and predicted profiles.

Final conclusions
For the main objective of WP4, the conclusion is obvious: There are a tremendous number of
chloride ingress models available and most of them utilize test results as input data. Test
results from any test method will have one or several models that require that particular data
as input parameters.

However, a very relevant conclusion is also: Most of these models are still not very accurate
in predicting reality.

The main drawback of the models is the description of the boundary conditions, i.e. the
chloride content in the exposed surface in a certain environment. Especially the increasing
chloride content with time is still not understood. Most models also need data from field
exposure to express the boundary conditions. Results from a test method are not reliable.

The time-dependency of chloride diffusion coefficients is still not well understood and
obviously described in very different ways in different models. This time-dependency is not
the result of any test method.

Standardizing the test methods proposed by the project ChlorTest is a large step forward to
improve the prediction of chloride ingress. The test methods only will, however, not be the
complete answer to the required information for accurate predictions.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 107 of 112
WP4 REPORT – Modelling of Chloride Ingress

REFERENCES
Alisa, M.; Andrade, C.; Gehlen, C.; Rodriguez, J.; Vogels, R.: Modelling of Degradation.
Brussels : Brite-Euram, 1998. Project No. BE95-1347, R 4/5, 174 pp.

Andrade, C. (2004), Calculation of initiation and propagation periods of service-life of


reinforcements by using the electrical resistivity, International Symposium on Advances in
Concrete through Science and Eng., RILEM Symposium, March 22-24, Evanston (Illinois,
USA), 2004.

Arliguie, G. (2003)Note de cours sur Durabilité des bétons, DEA Génie Civil, l’Institut
National des Sciences Appliquées de Toulouse, France, 2003

Bamforth P.B.(1993), Concrete classifications for R.C. structures exposed to marine and
other salt-loaded environments, Proceedings of Structural Faults and Repair, Edimburgh,
1993.

Bamforth, P. B.: (2004) Enhanching reinforced concrete durability, Concrete Society


Technical Report no 61, UK

Bjergovic, D.; Krstic, V.; Mikulic, D.; Ukrinczyk, V.(1994): C-D-c-t diagrams for practical
design of concrete durability parameters. Department of Materials, Faculty of Civil
Engineering. University of Zagreb.

Boddy, A., Bentz, E., Thomas, M.D.A. and Hooton, R.D. (1999) An overview and sensitivity
study of a multimechanistic chloride transport model. Cement and Concrete Research,
Volume 29, Issue 6, June 1999, Pages 827-837

Buenfeld, N.R.; Shurafa-Daoudi M.-T.; McLoughlin I.M.(1995): Chloride transport due to


wick action in concrete. Paper presented at the RILEM International Workshop on Chloride
Penetration into Concrete, October 15-18, Saint-Rémy-les-Chevreuse, France.

Collepardi, M.; Marcialis, A.; Turriziani, R.(1970): The kinetics of chloride ions penetration
in concrete (in Italian). Il Cemento, Vol. 67, pp. 157-164.

Crank, J.C. (1975), The mathematics of diffusion, 2nd edition, Oxford University Press.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 108 of 112
WP4 REPORT – Modelling of Chloride Ingress

Denarié, E., Conciatori, D. & Brühwiler, E. (2003), Effect of micro climate on chloride
penetration into reinforced concrete 6th CANMET/ACI International Conference on
Durability of Concrete, Tessaloniki, Greece, 2003.

Francy, O. (1998) Modélisation de la pénétration des ions chlorures dans les mortier
partiellement saturés en eau. These du doctorat, LMDC, Université Paul Sabatier, Toulouse
1998

Frederiksen, J. M., L.-O. Nilsson, P. Sandberg, E. Poulsen, Tang L. & A. Andersen A system
for estimation of chloride ingress into concrete. Theoretical background. HETEK. Danish
Road Directorate Report No 83 1997.

Frederiksen J.M., Geiker M.(2000), On an empirical model for estimation of chloride ingress
into concrete, Proceedings of the 2 nd International RILEM Workshop "Testing and
Modelling Chloride Ingress into Concrete", sept. 11-12, 2000, Paris, France (Ed. by C.
Andrade & J. Kropp, RILEM, 2000), pp 355-371.

Gehlen, C.; Ludwig, H.-M (1999).: Compliance Testing for Probabilistic Design Purposes:
Brussels: Brite-Euram,. Project No. BE95-1347, R 8, 114 pp.

Gehlen, C. (2000), Probabilistische Lebensdauerbemessung von Stahlbetonbauwerken –


Zuverlässigkeitsbetrachtungen zur wirksamen Vermeidung von Bewehrungskorrosion, Heft
510 DafStb, Germany

Gulikers, J. (2004), Critical evaluation of service life models applied on an existing marine
concrete structure. NORECON Seminar 2004: Repair and Maintenance of Concrete
Structures, Copenhagen April 19-20, 2004

Heinfling G., Courtois A., Hornet P., (1998) Application des méthodes probabilistes à
l'analyse du comportement des structures en béton viellisantes: une approche industrielle,
Seconde Conférence Nationale "Fiabilité des matériaux et des structures", France, 23-24
Novembre 1998

Hornet et coll., Hornet P., Dupas P., Schneiter J-R., (1998) Sensitivity Analyses on the
Behavior of a Pressure Vessel during a Severe Accident using Probabilistic Methods, Severe
Accidents and Topics in the NESC Project, ASME, PVP-Vol.362, pp 123-129, 1998

Houdusse O., Marchand J., Hornain H.(1998), Prédiction de l’évolution de la concentration


des ions présents dans les matériaux cimentaires : modélisation numérique du transport
d’ions par la méthode des éléments finis, Proceedings of the 1 st International Meeting
"Material Science and Concrete Properties", march 5-6, 1998, Toulouse, France, (LMDC,
1998), pp 151-158.

Houdusse O., Hornain H., Martinet G. (2000), Prediction of long-term durability of Vasco de
Gama Bridge in Lisbon, Proceedings of the 5 th CANMET/ACI International Conference on
Durability of Concrete, June 4-9, 2000, Barcelona, Spain, SP-192 (Ed. by V.M. Malhotra,
ACI, 2000), vol. II, SP 192-20.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 109 of 112
WP4 REPORT – Modelling of Chloride Ingress

Johannesson, B. (2000), Transport and sorption phenomena in concrete and others porous
media, PhD thesis, report TVBM-1019, div. of Building Materials, Lund Institute of
Technology, Lund, Sweden

Khitab, A., Lorente, S., Ollivier, J.P.(2004), Chloride diffusion through saturated concrete:
numerical and experimental results, Advances in Concrete through Science and Eng.,
RILEM Symposium, March 22-24, Evanston (Illinois, USA), 2004.

Lee N.P., Christholm D.H.(2000), Impediments to the application of marine concrete


durability models in design, Proceedings of the 2 nd International RILEM Workshop
"Testing and Modelling Chloride Ingress into Concrete", sept. 11-12, 2000, Paris, France (Ed.
by C. Andrade & J. Kropp, RILEM, 2000, pp 383-393.

Li, L.Y. and Page, C.L. (1998), Modelling of electrochemical chloride extraction from
concrete: influence of ionic activity coefficients, Computational Materials Science, 9, 1998,
303-308.

Lindvall, A. (2003), Environmental actions on concrete exposed in marine and road


environments and its response – consequences for the initiation of chloride induced
reinforcement corrosion. PhD thesis, publication P-03:2, dept. of Building Technology –
Building Materials, Chalmers University of Technology, Göteborg, Sweden

Maage M., Helland S. (1991) Quality Inspection of “Shore Approach” High strength
concrete. Proceedings Second CANMET/ACI International Conference on Durability of
Concrete, Montreal 1991, ACI SP 126, Detroit USA
.
Maage, M.; Poulsen, E.; Vennesland, Ø.; Carlsen, J.E.(1995): Service life model for concrete
structures exposed to marine environment  initiation period. LIGHTCON Report No. 2.4,
STF70 A94082 SINTEF, Trondheim, Norway.

Maage M., Helland S. Carlsen J. E. (1999) Chloride penetration into concrete with light
weight aggregates. Brite EuRam project BE96-3942, CUR P.O.Box 420, NL- 2800 AK
Gouda, The Netherlands, May 1999. This can also be found as report R3 in
http://www.sintef.no/static/BM/projects/EuroLightCon/Rapportoversikt.htm

Madsen et coll., Madsen H.O., Krenk S., Lind N.C., (1986) Method of structural safety,
Prentice-Hall Inc, ISBN 0-13-579475-7, 1986

Mangat, P.S., Molloy, B.T.(1994): Predicting of long term chloride concentration in


concrete. Materials and Structures, Vol. 27, pp. 338-346.

Marchand J., Samson E., Maltais Y., Lee R.J., Sahu S.(2002), Predicting the performance of
concrete structures exposed to chemically aggressive environment. Field validation, Concrete
science and engineering, vol. 4, n° 15, included in Materials and structures, vol. 35, n° 254,
Dec. 2002, pp 623-631.

Masi M., Colella D., Radaelli G., Bertolini L.(1997), Simulation of chloride penetration in
cement-based materials, Cement and concrete research, vol. 27, n° 10, 1997, pp 1591-1601.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 110 of 112
WP4 REPORT – Modelling of Chloride Ingress

McLoughlin I.M. (1997), Modelling of chloride and moisture transport in concrete, PhD
thesis, dept. of Civil Engineering, Imperial College, London 1997.

Meijers, S.J.H. (2003), Computational modelling of chloride Ingress in concrete, PhD thesis,
Delft University, The Netherlands.

Mejlbro, L.(1996): The complete solution of Fick’s second law of diffusion with time-
dependent diffusion coefficient and surface concentration. Durability of marine concrete
structures. CEMENTA, Danderyd. Sweden

Mohammed, T. U., Yamaji, T. Hamada, H. (2002) Microstructures and interfaces in


concrete after 15 years of exposure in tidal environment. ACI Materials Journal, Vol 99, No
4 – July-August 2002

Nilsson, L.O.; Ollivier, J.P. (editors) (1995): Chloride penetration into concrete. Papers
presented at the RILEM “International Workshop on Chloride Penetration into Concrete”,
October 15-18, 1995, Saint Rémy les Chevreuse, France.

Nilsson, L-O, E Poulsen, P Sandberg, H E Sörensen, O Klinghoffer (1996) Chloride


penetration into concrete. State of the Art. HETEK. Danish Road Directorate Report No 53
1996.

Nilsson, L.-O.(1997), A model for convection of chloride, chapter 7 in Frederiksen et al


(1997)

Nilsson, L-O (2000) A numerical model for combined diffusion and convection of chloride in
non-saturated concrete. Proceedings of 2nd International Workshop on Testing and
Modelling the Chloride Ingress into Concrete, 11-12 September 2000, Paris

Nilsson, L-O (2001) Prediction models for chloride ingress and corrosion initiation in
concrete structures, Nordic Mini Seminar & fib TG 5.5 meeting, Göteborg, May 22-23,
2001. Publication P-01:6, Dept. Building Materials, Chalmers University of Technology,
December 2001

Nilsson, L.-O. (2002) On the uncertainty of service-life models for reinforced marine
concrete structures, International RILEM workshop Life Prediction and Age Management of
Concrete Structures, Cannes, 16-17 October 2002

Nilsson, L-O (2002) Concepts In Chloride Ingress Modelling Key note paper at 3rd
International RILEM workshop on Testing And Modelling Chloride Ingress Into Concrete,
Madrid, 9-10 September 2002.

Nguyen Xuan Son, (2004) L'approche multi-espèces pour la description de la diffusion des
chloures, Mémoire bibliographique de DEA Génie Civil, LMDC, INSA de Toulouse, France,
2004
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 111 of 112
WP4 REPORT – Modelling of Chloride Ingress

Nugue, F. (2002) Recherche d’une méthode rapide de détermination du coefficient de


diffusion en milieu cimentaire saturé, Thèse de doctorat de l’Institut National des Sciences
Appliquées de Toulouse, France, 2002

Olesen R. (1992) PROBAN User's Manual, Technical report Det Norske Veritas, 1992

Ollivier, J.-P., Carcassès, M., Bigas, J.-P. & Truc, O. (2002) Diffusion des chloures dans le
béton saturé, Revue française de génie civil, Volume 6 - No.2/2002, pp 227 à 250

Petre-Lazar I., Heinfling G., Marchand J., Gérard B. (2000), Application of probabilistic
methods to analysis of behavior of reinforced concrete structures affected by steel corrosion,
Proceedings of the 5 th CANMET/ACI International Conference on Durability of Concrete,
June 4-9, 2000, Barcelona, Spain, SP-192 (Ed. by V.M. Malhotra, ACI, 2000), pp 557-572.

Petre-Lazar, I. (2000) Evaluation du comportement en service des ouvrages en béton armé


soumis à la corrosion des aciers, Thèse de doctorat de l’Université Laval de Québec, Canada,
2000

Petre-Lazar, I., L. Abdou, C. Franco, I. Sadri (2003) THI - A Physical Model for Estimating
the Coupled Transport of Heat, Moisture and Chloride Ions in Concrete. 2nd International
RILEM Workshop on Life Prediction and Aging Management of Concrete Structures, Paris,
5-6 May 2003, France

Poulsen, E.(1990): Chlorides and 100 years of service-life (in Danish) Proceedings of “Dansk
Betondag 1990”. Dansk Betonforening, Publikation No. 36. Copenhagen, Denmark.

Poulsen, E.(1993): On a model of chloride ingress into concrete having time dependent
diffusion coefficient. Proceedings of the Nordic Miniseminar in Gothenburg. Chalmers
University of Technology. Gothenburg, Sweden. (Editor: L-O Nilsson). Report P-93:1.

Poulsen, E.(1995): The LIGHTCON service life model improved at the RILEM workshop in
St-Rémy (in Danish). Internal AECnote. AEC Consulting Engineers (LtD) A/S, Vedbæk,
Denmark.

Samson E., Marchand J.(1999), Numerical solution of the extended Nernst-Planck Model,
Journal of Colloid and Interface Science, 215, 1999, pp 1-8.

Samson E., Marchand J., Robert L., Bournazel J.P. (1999), Modeling the mechanisms of ion
diffusion in porous media, International journal for numerical methods in engineering, 46,
1999, pp 2043-2060.

Samson E., Lemaire G., Marchand J., Beaudoin J.(1999), Modeling chemical activity effects
in strong ionic solutions, Computational material science, 15, 1999, pp 285-294.

Siemes, A.J.M.; Gehlen, C.; Lindvall, A.; Arteaga, A.; Ludwig, H.-M.: Statistical
Quantification of the Variables in the Limit State Functions. Summary: Brussels: Brite-
Euram, 1999. Project No. BE95-1347, Draft R 9, 106 pp.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 112 of 112
WP4 REPORT – Modelling of Chloride Ingress

Tang, L.(1996) Chloride Transport in Concrete - Measurement and Prediction, Gothenburg,


Chalmers University of Technology, 1996

Tang, L. (2005) Engineering expression of the ClinConc model for prediction of free and
total chloride ingress in submerged marine concrete, submitted to Cement and Concrete
Research.

Truc, O. (2000), Prediction of Chloride Penetration into Saturated Concrete. Multi-Species


Approach, Thèse de doctorat de l'Université de technologie de Chalmers et de l'Institut
National des Sciences Appliquées, Göteborg et Toulouse, 2000, 173 p

Tuutti, K. Corrosion of steel in concrete. Research report No.4.82Swedish Cement and


Concrete Research Institute (CBI), Stockholm 1982

Visser, J.H.M., Gaal, G.C.M., and de Rooij, M.R (2002), Time dependency of chloride
diffusion coefficients in concrete, 3rd RILEM Workshop Testing and Modelling Chloride
Ingress into Concrete, Madrid, September 9-10, 2002, to be published
CHLORTEST – EU Funded Research Project under 5FP GROWTH Programme

Resistance of concrete to chloride ingress


– From laboratory tests to in-field performance

WP5 REPORT – FINAL EVALUATION


OF TEST METHODS

Deliverables D16-19
CONTRACT N°: G6RD-CT-2002-00855
PROJECT N°: GRD1-2002-71808
ACRONYM: CHLORTEST
DURATION: January 2003 – December 2005
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 2 of 38
WP5 REPORT – Final Evaluation of Test Methods

PROJECT
COORDINATOR: SP Swedish National Testing and Research Institute (SP) S

PARTNERS: Institute of Construction Sciences “Eduardo Torroja” (IETcc) E


University of Alicante (UoA) E
Chalmers University of Technology (Chalmers) S
Skanska Norge AS (Selmer) NO
Swedish National Road Administration (SNRA) S
Electricité de France (EDF) F
Netherlands Organisation for Applied Scientific Research (TNO) NL
Hochschule Bremen (HSB) D
Slovenian National Building and Civil Engineering Institute (ZAG) SI
Queens University Belfast (QUB) UK
Laboratório Nacional de Engenharia Civil (LNEC) P
Icelandic Building Research Institute (IBRI) IS
National Institute of Applied Science (INSA) F
Laboratoire Central des Ponts et Chaussées (LCPC) F
Valenciana de Cementos, S.A. CEMEX (VCLC) E
Lund Institute of Technology (LTH) S

ACKNOWLEDGEMENT:
The present document is deliverables of Workpackage 5 – “Final Evaluation of Test
Methods”. The consortium members SP, IETcc, UoA, Chalmers, EDF, TNO,, HSB, ZAG,
QUB, LNEC, IBRI, INSA, LCPC, VCLC and LTH were involved in the work of this part of the
project. The work was led by SP, assisted by IETcc.

This document was prepared by Tang Luping (SP)

FURTHER INFORMATION:
Regarding this document: Regarding the project in general
Dr Tang Luping Dr Tang Luping
SP Swedish National Testing and Research SP Swedish National Testing and Research
Institute Institute
Box 857 Box 857
S-501 15 BORÅS, Sweden S-501 15 BORÅS, Sweden
Tel. +46-33 165138; Fax: +46-33 134516 Tel. +46-33 165138; Fax: +46-33 134516
e-mail: tang.luping@sp.se e-mail: tang.luping@sp.se
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 3 of 38
WP5 REPORT – Final Evaluation of Test Methods

TABLE OF CONTENTS
Page

1 INTRODUCTION 5

2 PARTICIPATING LABORATORIES 5

3 CONCRETE 6

4 TEST PROCEDURES AND SPECIMENS 7

5 TEST RESULTS AND PRECISION ANALYSIS 9


5.1 Primary examination 9
5.2 Consistency Test 9
5.3 Criteria for rejection of data 9
5.4 Precision analysis and results 9
5.5 Dependency analysis 24

6 DISCUSSIONS 28
6.1 Comparison of the precision results 28
6.2 Comparison of the mean values 30
6.3 Relationships between various transport parameters 32

7 CONCLUDING REMARKS AND SUGGESTIONS 34


7.1 Census opinions 34
7.2 Precision of various methods 35
7.3 Suggestions 35

REFERENCES 36

Appendix 1: Description of Method M6 (by M. Castellote, C. Andrade, and C. Alonso)

Appendix 2: Description of Method R1 (by C. Andrade)

Appendix 3: Summary of the results from individual laboratories

Appendix 4: Received answers to the questionnaire


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 4 of 38
WP5 REPORT – Final Evaluation of Test Methods
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 5 of 38
WP5 REPORT – Final Evaluation of Test Methods

1 INTRODUCTION
The objectives of this workpackage are

• Evaluating four test methods selected based on the results from WP2 [1], and
• Producing precision data of the selected test methods according to ISO 5725.

In this workpackage 15 laboratories (14 contracted and 1 voluntary) participated in the inter-
laboratory tests for evaluation of the following four selected test methods:

• D2 (NT BUILD 443), as for non-steady state diffusion;


• M4 (NT BUILD 492), as for non-steady state migration;
• M6 (steady-state migration test based on the conductivity measurement), as for
steady-state as well as non-steady state migration; and
• R1 (resistivity test), as for rapid indication of transport property.

Six types of concrete manufactured with the combination of four types of binder (PC, and PC
blended with silica fume, slag and fly ash) were used in the evaluation.

This report presents the results from the inter-laboratory test and the precision analysis.

2 PARTICIPATING LABORATORIES
The laboratories participated in the inter-laboratory test for various test methods are listed in
Table 2.1.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 6 of 38
WP5 REPORT – Final Evaluation of Test Methods

Table 2.1 – Participating laboratories for various test methods.


No. Partner D2 M4 M6 R1
P1 SP X X X X
P2 IETcc X X X X
P3 UoA X X X X
P4 Chalmers X X - X
P7 EDF - X X (failed) X
P8 TNO - X X X
P9 HSB X X X X
P10 ZAG Participated but no result reported
P11 QUB X X X X
P12 LNEC X X X X
P13 IBRI X X X
P14 INSA X X X X
P15 LCPC - X X X
P16 VCLC X - X X
P17* LTH - X (partly) - -
X: participated; * Voluntary.

3 CONCRETE
Six types of concrete were used in the inter-laboratory test. The mixture proportions of
concrete are listed in Table 3.1. Four types of binder were tested. Swedish natural sand and
gravel were used as fine and coarse aggregates, respectively. Concrete slabs of size
1200×800×200 mm were cast at SP in May 2004 and cured without demoulding under moist
condition (covered with thick plastic foils) at the room temperature for at least four weeks,
and afterwards the moist condition was not assured even though the slabs were still covered
with thick plastic foils. After about five months storage in the laboratory the cores of size
∅100×200 mm (∅75×200 mm for participant No. 2) were taken from the slabs. Three cores
per concrete type were randomly selected (except for participant No. 2 who needed ∅75 mm
cores) and distributed to each participating laboratory in December 2004.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 7 of 38
WP5 REPORT – Final Evaluation of Test Methods

Table 3.1 – Mixture proportions of concrete used in the inter-laboratory test.


Material PC50 PC42 PC35 SF42 FA42 SL42
Cement type Swedish Norwegian Dutch CEM III/B
CEM I 42.5 N BV/SR/LA CEM II/A-V 42.5 R 42.5 LH HS
(∼18% fly ash) (∼70% slag)
Cement 400 420 450 389.5 410 410
Silica fume 20.5
(ELEKM)
Water 200 176.4 157.5 172.2 172.2 172.2
Sand (0-8 mm) 920 926 904 897 901 901
Gravel 816 855 904 897 901 901
(10-15 mm)
SP (CemPlux) 0.5 1.0 0.5 0.5 0.5
% of binder
w/b 0.5 0.42 0.35 0.42 0.42 0.42

4 TEST PROCEDURES AND SPECIMENS


Methods D2 and M4 are already the Nordic standard, NT BUILD 443 and 492, respectively.
The descriptions of Methods M6 and R1 are attached in Appendixes 1 and 2. To share the
specimens and preconditioning procedures, Method R1 was carried out in combination with
Methods M4 and M6. It should be noticed that in M4, saturated lime water is used for
vacuum saturation, while in M6 demineralised water is used for vacuum saturation.

The test specimens for various methods were cut at individual laboratories according to the
schematic shown in Figure 4.1. It should be noticed that, owing to insufficient vibration
under casting, some defects (air voids or cavities) were found in some cores, as shown in
Figure 4.2. In order to keep away the visual defects, the schematic in Figure 4.1 may not
exactly be followed in some laboratories.

Depending on laboratories, the unavoidable visual defects in specimens were handled in


different ways. Some of laboratories did nothing while some used silicon or other sealant to
fill up the voids or cavities. The effect of defects will be discussed later.

The inter-laboratory tests started in January 2005. Most of the participating laboratories
finished the tests by the end of March and some of them by the end of June.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 8 of 38
WP5 REPORT – Final Evaluation of Test Methods

Cl exposure surface
For M6
Trowelled
surface
The rest for initial Cl according to D2
Cl exposure surface (one specimen is enough)
For M4
20 ± 2 mm for M6 and R1 (3 specimens from 3 cores)
Cl exposure surface
For D2 50 ± 2 mm for M4 and R1 (3 specimens from 3 cores)

60-70 mm for D2 (3 specimens from 3 cores)

Figure 4.1 – Schematic for specimen cutting.

Figure 4.2 – Example of the defects in some specimens (by courtesy of UoA).
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 9 of 38
WP5 REPORT – Final Evaluation of Test Methods

5 TEST RESULTS AND PRECISION ANALYSIS


The test results reported by individual laboratories are summarised in Appendix 3. These
results were double-checked by each laboratory to assure the correctness.

The precision analysis was carried out according to ISO 5725:94.

5.1 Primary examination


The raw data were visually examined for possible abnormal values. Two values in the
parameter Dns from Method M6 seem significantly abnormal, that is,

Lab 11: PC35 test 2, the value 105.73 in (24.07, 105.73, 19.8)
Lab 11: SL42 test 2, the value 741.11 in (20.57, 741.11, 9.88)

These two values were excluded before the consistency test.

5.2 Consistency Test


Similar to WP2, two methods as described in ISO 5725-2:94 were used for the consistency
test, that is, Mandel’s k-statistic and Cochran’s test for within-laboratory consistency, and
Mandel’s h-statistic and Grubbs’ test (single only) for between-laboratory consistency.
According to the standard, the significant level α = 0.01 was used for the criteria of outlier
and α = 0.05 for the criteria of straggler.

The consistency test was carried out only once, but if a laboratory was excluded as outlier,
that is, all the data from this laboratory were rejected due to many outliers, the consistency
test was carried out again.

5.3 Criteria for rejection of data


To assure the proper rejection of data, the consistency test results from both methods should
be consistent. Therefore, to reject one set of data, it should be classified as outlier by both
Mandel’s k -statistic and Cochran’s tests or by both Mandel’s h -statistic and Grubbs’ (single)
tests.

5.4 Precision analysis and results


5.4.1 Method D2
From both Mandel’s h -statistic and Grubbs’ tests it was found that most of the data from Lab
16 were outliers, as shown in Figure 5.1 and Table 5.1. Therefore, this laboratory was
rejected as outlier, and the consistency test was carried out again after rejection of Lab 16.
The results from the second consistency test are shown in Figure 5.2 and Tables 5.2 and 5.3.
Based on Figure 5.2 and Table 5.2, two outliers (PC35 and FA42 from Lab 13) were
excluded from the precision calculation. Since the value Cs is accompanied with Dns and the
value K is mathematically derived from Dns, the same rejection as applied to Dns was applied
to both Cs and K without further consistency test. The precision results are shown in Tables
5.4-5.6.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 10 of 38
WP5 REPORT – Final Evaluation of Test Methods

Within-laboratory Consistency Statistic


According to ISO 5725

PC50
Mandel's k -statistic

PC42
2
PC35
SF42
FA42
SL42
1 1% significance for PC50
5% significance for PC50

0
Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab
1 2 3 4 7 8 9 10 11 12 13 14 15 16 17

Laboratory

Between-laboratory Consistency Statistic


According to ISO 5725

2
PC50
Mandel's h -statistic

PC42
1
PC35
SF42
0
FA42
SL42
-1 1% significance for PC50
5% significance for PC50
-2

-3
Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab
1 2 3 4 7 8 9 10 11 12 13 14 15 16 17

Laboratory

Figure 5.1 - Mandel’s k-statistic and h-statistic of value Dns from Method D2.

Table 5. 1 – Grubbs’ test of value Dns from Method D2


Level j PC50 PC42 PC35 SF42 FA42 SL42
Valid lab p 8 9 8 9 10 9
Single G cr(1%) 2.274 2.387 2.274 2.387 2.482 2.387
Single G cr(5%) 2.126 2.215 2.126 2.215 2.29 2.215
Single low G 1 0.620 1.132 0.619 0.430 0.389 0.641
Single high G p 2.383 2.327 2.449 2.663 2.843 2.610
Single-low outlier criterion Correct Correct Correct Correct Correct Correct
Single-low outlier Lab
Single-high outlier criterion Outlier Straggler Outlier Outlier Outlier Outlier
Single-high outlier Lab Lab 16 Lab 16 Lab 16 Lab 16 Lab 16 Lab 16
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 11 of 38
WP5 REPORT – Final Evaluation of Test Methods

Within-laboratory Consistency Statistic


According to ISO 5725

PC50
Mandel's k -statistic

PC42
2
PC35
SF42
FA42
SL42
1 1% significance for FA42
5% significance for FA42

0
Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab
1 2 3 4 7 8 9 10 11 12 13 14 15 16 17

Laboratory

Between-laboratory Consistency Statistic


According to ISO 5725

2
PC50
Mandel's h -statistic

PC42
1
PC35
SF42
0
FA42
SL42
-1 1% significance for FA42
5% significance for FA42
-2

-3
Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab
1 2 3 4 7 8 9 10 11 12 13 14 15 16 17

Laboratory

Figure 5.2 - Mandel’s k-statistic and h-statistic of value Dns from Method D2, after rejection
of all the data from Lab 16.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 12 of 38
WP5 REPORT – Final Evaluation of Test Methods

Table 5.2 - Cochran's test results for value Dns from Method D2, after rejection of all the data
from Lab 16
Level j PC50 PC42 PC35 SF42 FA42 SL42
Valid lab p 7 8 7 8 9 8
Mean n 3 3 3 3 3 3
C cr(1%) 0.664 0.615 0.664 0.615 0.573 0.615
C cr(5%) 0.561 0.516 0.561 0.516 0.478 0.516
Cochran's test statistic C 0.615 0.492 0.720 0.373 0.609 0.371
Outlier criterion Straggler Correct Outlier Correct Outlier Correct
Outlier Lab Lab 13 Lab 13 Lab 13

Table 5.3 - Grubbs’ test results for value Dns from Method D2, after rejection of all the data
from Lab 16
Level j PC50 PC42 PC35 SF42 FA42 SL42
Valid lab p 7 8 7 8 9 8
Single G cr(1%) 2.139 2.274 2.139 2.274 2.387 2.274
Single G cr(5%) 2.02 2.126 2.02 2.126 2.215 2.126
Single low G 1 0.961 1.610 1.732 1.714 1.487 1.438
Single high G p 1.292 1.705 1.259 0.834 0.927 1.675
Single-low outlier criterion Correct Correct Correct Correct Correct Correct
Single-low outlier Lab
Single-high outlier criterion Correct Correct Correct Correct Correct Correct
Single-high outlier Lab

Table 5.4 – Results of precision analysis for value Dns [×10-12 m2/s] from Method D2.
Level j PC50 PC42 PC35 SF42 FA42 SL42
Number of valid lab p 7 8 6 8 8 8
General mean m 17.54 15.65 4.66 4.073 1.455 1.498
2
Repeatability variance s r 16.55 24.205 0.33918 0.21113 0.05042 0.15886
2
Between-lab variance s L 15.545 6.844 1.13061 0.96741 0.02827 0.04298
2
Reproducibility variance s R 32.095 31.049 1.46979 1.17854 0.07869 0.20184
Repeatability std. dev. s r 4.07 4.92 0.582 0.459 0.225 0.399
Reproducibility std. dev. s R 5.67 5.57 1.212 1.086 0.281 0.449
Repeatability COV(s r) 23.2% 31.4% 12.5% 11.3% 15.5% 26.6%
Reproducibility COV(s R) 32.3% 35.6% 26.0% 26.7% 19.3% 30.0%
γ = s R/s r 1.39 1.13 2.08 2.37 1.25 1.13
Repeatability limit r = 2.8s r 11.4 13.78 1.63 1.285 0.63 1.117
Reproducibility limit R = 2.8s R 15.9 15.6 3.39 3.04 0.79 1.26
Number of excluded outliers 1 1 2 1 2 1
Outlier type G G C, G G C, G G
Outlier laboratories Lab 16 Lab 16 Lab 13, 16 Lab 16 Lab 13, 16 Lab 16
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 13 of 38
WP5 REPORT – Final Evaluation of Test Methods

Table 5.5 – Results of precision analysis for value Cs [mass% of sample] from Method D2.
Level j PC50 PC42 PC35 SF42 FA42 SL42
Number of valid lab p 7 8 6 8 8 8
General mean m 0.775 0.7614 0.7834 0.8619 1.0619 0.7046
2
Repeatability variance s r 0.0083425 0.0202831 0.0127302 0.0104277 0.0479599 0.0416703
2
Between-lab variance s L 0.0111602 0 0.0234493 0.0222022 0.0104164 0
2
Reproducibility variance s R 0.0195027 0.0202831 0.0361795 0.0326299 0.0583763 0.0416703
Repeatability std. dev. s r 0.0913 0.1424 0.1128 0.1021 0.219 0.2041
Reproducibility std. dev. s R 0.1397 0.1424 0.1902 0.1806 0.2416 0.2041
Repeatability COV(s r) 11.8% 18.7% 14.4% 11.8% 20.6% 29.0%
Reproducibility COV(s R) 18.0% 18.7% 24.3% 21.0% 22.8% 29.0%
γ = s R/s r 1.53 1.00 1.69 1.77 1.10 1.00
Repeatability limit r = 2.8s r 0.2556 0.3987 0.3158 0.2859 0.6132 0.5715
Reproducibility limit R = 2.8s R 0.391 0.399 0.533 0.506 0.676 0.571
Number of excluded outliers 1 1 2 1 2 1
Outlier type G G C, G G C, G G
Outlier laboratories Lab 16 Lab 16 Lab 13, 16 Lab 16 Lab 13, 16 Lab 16

Table 5.6 – Results of precision analysis for value K [mm/√yr] from Method D2.
Level j Level 1 Level 2 Level 3 Level 4 Level 5 Level 6
Number of valid lab p 7 8 6 8 8 8
General mean m 63.2 59.55 34.08 32.1 20.92 18.74
2
Repeatability variance s r 26.796 54.476 6.437 2.184 5.802 3.554
2
Between-lab variance s L 63.108 37.124 0.556 9.915 5.723 4.775
2
Reproducibility variance s R 89.904 91.6 6.993 12.099 11.525 8.329
Repeatability std. dev. s r 5.18 7.38 2.54 1.48 2.41 1.89
Reproducibility std. dev. s R 9.48 9.57 2.64 3.48 3.39 2.89
Repeatability COV(s r) 8.2% 12.4% 7.5% 4.6% 11.5% 10.1%
Reproducibility COV(s R) 15.0% 16.1% 7.7% 10.8% 16.2% 15.4%
γ = s R/s r 1.83 1.30 1.04 2.35 1.41 1.53
Repeatability limit r = 2.8s r 14.5 20.66 7.11 4.14 6.75 5.29
Reproducibility limit R = 2.8s R 26.5 26.8 7.4 9.7 9.5 8.1
Number of excluded outliers 1 1 2 1 2 1
Outlier type h1 h1 k2, h1 h1 k2, h1 h1
Outlier laboratories Lab 16 Lab 16 Lab 13, 16 Lab 16 Lab 13, 16 Lab 16
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 14 of 38
WP5 REPORT – Final Evaluation of Test Methods

5.4.2 Method M4
The results from the consistency test are shown in Figure 5.3 and Tables 5.7 and 5.8. Only
one outlier (FA42 from Lab 15) was excluded according to the criteria. The precision results
are shown in Table 5.9.

Within-laboratory Consistency Statistic


According to ISO 5725

PC50
Mandel's k -statistic

PC42
2
PC35
SF42
FA42
SL42
1 1% significance for FA42
5% significance for FA42

0
Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab
1 2 3 4 7 8 9 10 11 12 13 14 15 16 17

Laboratory

Between-laboratory Consistency Statistic


According to ISO 5725

2
PC50
Mandel's h -statistic

PC42
1
PC35
SF42
0
FA42
SL42
-1 1% significance for FA42
5% significance for FA42
-2

-3
Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab
1 2 3 4 7 8 9 10 11 12 13 14 15 16 17

Laboratory

Figure 5.3 - Mandel’s k-statistic and h-statistic of value Dns from Method M4.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 15 of 38
WP5 REPORT – Final Evaluation of Test Methods

Table 5.7 - Cochran's test results for value Dns from Method M4
Level j PC50 PC42 PC35 SF42 FA42 SL42
Valid lab p 12 12 13 13 13 13
Mean n 3 3 3 3 3 3
C cr(1%) 0.475 0.475 0.450 0.450 0.450 0.450
C cr(5%) 0.392 0.392 0.371 0.371 0.371 0.371
Cochran's test statistic C 0.219 0.311 0.301 0.205 0.604 0.350
Outlier criterion Correct Correct Correct Correct Outlier Correct
Outlier Lab Lab 15

Table 5.8 – Grubbs’ test results for value Dns from Method M4
Level j PC50 PC42 PC35 SF42 FA42 SL42
Valid lab p 12 12 13 13 13 13
Single G cr(1%) 2.636 2.636 2.699 2.699 2.699 2.699
Single G cr(5%) 2.412 2.412 2.462 2.462 2.462 2.462
Single low G 1 1.909 1.607 1.378 1.656 1.574 1.555
Single high G p 1.539 1.328 1.856 1.513 1.586 2.205
Single-low outlier criterion Correct Correct Correct Correct Correct Correct
Single-low outlier Lab
Single-high outlier criterion Correct Correct Correct Correct Correct Correct
Single-high outlier Lab

Table 5.9 – Results of precision analysis for value Dns [×10-12 m2/s] from Method M4.
Level j PC50 PC42 PC35 SF42 FA42 SL42
Number of valid lab p 12 12 13 13 12 13
General mean m 15.43 14.67 5.944 7.441 2.313 1.765
2
Repeatability variance s r 1.793 3.854 0.54578 0.8154 0.26389 0.15412
2
Between-lab variance s L 4.865 6.472 0.42539 1.75785 0.25753 0.19225
2
Reproducibility variance s R 6.658 10.326 0.97117 2.57325 0.52142 0.34637
Repeatability std. dev. s r 1.34 1.96 0.739 0.903 0.514 0.393
Reproducibility std. dev. s R 2.58 3.21 0.985 1.604 0.722 0.589
Repeatability COV(s r) 8.7% 13.4% 12.4% 12.1% 22.2% 22.3%
Reproducibility COV(s R) 16.7% 21.9% 16.6% 21.6% 31.2% 33.4%
γ = s R/s r 1.93 1.64 1.33 1.78 1.40 1.50
Repeatability limit r = 2.8s r 3.75 5.49 2.069 2.528 1.439 1.1
Reproducibility limit R = 2.8s R 7.2 9 2.76 4.49 2.02 1.65
Number of excluded outliers 0 0 0 0 1 0
Outlier type C
Outlier laboratories Lab 15
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 16 of 38
WP5 REPORT – Final Evaluation of Test Methods

5.4.3 Method M6
For the value Ds, the results from the consistency test are shown in Figure 5.4 and Tables
5.10 and 5.11. Three outliers (PC42 from Lab 11, FA42 and SL42 from Lab 16) were
excluded according to the criteria. The precision results are shown in Table 5.12.

Within-laboratory Consistency Statistic


According to ISO 5725

PC50
Mandel's k -statistic

PC42
2
PC35
SF42
FA42
SL42
1 1% significance for PC50
5% significance for PC50

0
Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab
1 2 3 4 7 8 9 10 11 12 13 14 15 16 17

Laboratory

Between-laboratory Consistency Statistic


According to ISO 5725

2
PC50
Mandel's h -statistic

PC42
1
PC35
SF42
0
FA42
SL42
-1 1% significance for PC50
5% significance for PC50
-2

-3
Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab
1 2 3 4 7 8 9 10 11 12 13 14 15 16 17

Laboratory

Figure 5.4 - Mandel’s k-statistic and h-statistic of value Ds from Method M6.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 17 of 38
WP5 REPORT – Final Evaluation of Test Methods

Table 5.10 - Cochran's test results for value Ds from Method M6


Level j PC50 PC42 PC35 SF42 FA42 SL42
Valid lab p 11 11 11 11 11 11
Mean n 3 3 3 3 3 3
C cr(1%) 0.504 0.504 0.504 0.504 0.504 0.504
C cr(5%) 0.417 0.417 0.417 0.417 0.417 0.417
Cochran's test statistic C 0.312 0.525 0.267 0.361 0.708 0.583
Outlier criterion Correct Outlier Correct Correct Outlier Outlier
Outlier Lab Lab 12 Lab 16 Lab 16

Table 5.11 - Grubbs’ test results for value Ds from Method M6


Level j PC50 PC42 PC35 SF42 FA42 SL42
Valid lab p 11 11 11 11 11 11
Single G cr(1%) 2.564 2.564 2.564 2.564 2.564 2.564
Single G cr(5%) 2.355 2.355 2.355 2.355 2.355 2.355
Single low G 1 1.059 1.231 1.169 1.197 1.078 1.167
Single high G p 2.334 1.311 1.717 1.723 1.967 2.021
Single-low outlier criterion Correct Correct Correct Correct Correct Correct
Single-low outlier Lab
Single-high outlier criterion Correct Correct Correct Correct Correct Correct
Single-high outlier Lab

Table 5.12 – Results of precision analysis for value Ds [×10-12 m2/s] from Method M6.
Level j PC50 PC42 PC35 SF42 FA42 SL42
Number of valid lab p 11 10 11 11 10 10
General mean m 2.859 2.051 1.377 1.624 0.59 0.372
2
Repeatability variance s r 0.73974 0.15301 0.07624 0.14165 0.01091 0.0067
2
Between-lab variance s L 6.52627 1.90748 1.02761 1.10256 0.20728 0.05509
2
Reproducibility variance s R 7.26601 2.06049 1.10385 1.24421 0.21819 0.06179
Repeatability std. dev. s r 0.86 0.391 0.276 0.376 0.104 0.082
Reproducibility std. dev. s R 2.696 1.435 1.051 1.115 0.467 0.249
Repeatability COV(s r) 30.1% 19.1% 20.0% 23.2% 17.6% 22.0%
Reproducibility COV(s R) 94.3% 70.0% 76.3% 68.7% 79.2% 66.9%
γ = s R/s r 3.13 3.67 3.81 2.97 4.49 3.04
Repeatability limit r = 2.8s r 2.408 1.095 0.773 1.053 0.291 0.23
Reproducibility limit R = 2.8s R 7.55 4.02 2.94 3.12 1.31 0.7
Number of excluded outliers 0 1 0 0 1 1
Outlier type C C C
Outlier laboratories Lab 12 Lab 16 Lab 16
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 18 of 38
WP5 REPORT – Final Evaluation of Test Methods

For the value Dns, the results from the consistency test are shown in Figure 5.5 and Tables
5.13 and 5.14. It should be noticed that two abnormal values from Lab 11 as mentioned in
Chapter 1 have primarily been excluded before the consistency test. Four outliers (PC50 and
PC35 from Lab 13, SF42 from Lab 15 and FA42 from Lab 8) were excluded according to the
criteria. The precision results are shown in Table 5.15.

Within-laboratory Consistency Statistic


According to ISO 5725

PC50
Mandel's k -statistic

PC42
2
PC35
SF42
FA42
SL42
1 1% significance for PC50
5% significance for PC50

0
Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab
1 2 3 4 7 8 9 10 11 12 13 14 15 16 17

Laboratory

Between-laboratory Consistency Statistic


According to ISO 5725

2
PC50
Mandel's h -statistic

PC42
1
PC35
SF42
0
FA42
SL42
-1 1% significance for PC50
5% significance for PC50
-2

-3
Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab
1 2 3 4 7 8 9 10 11 12 13 14 15 16 17

Laboratory

Figure 5.5 - Mandel’s k-statistic and h-statistic of value Dns from Method M6.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 19 of 38
WP5 REPORT – Final Evaluation of Test Methods

Table 5.13 - Cochran's test results for value Dns from Method M6
Level j PC50 PC42 PC35 SF42 FA42 SL42
Valid lab p 11 10 11 11 11 11
Mean n 3 3 3 3 3 3
C cr(1%) 0.504 0.536 0.504 0.504 0.504 0.504
C cr(5%) 0.417 0.445 0.417 0.417 0.417 0.417
Cochran's test statistic C 0.884 0.462 0.723 0.609 0.615 0.302
Outlier criterion Outlier Straggler Outlier Outlier Outlier Correct
Outlier Lab Lab 13 Lab 13 Lab 13 Lab 15 Lab 8

Table 5.14 - Grubbs’ test results for value Dns from Method M6
Level j PC50 PC42 PC35 SF42 FA42 SL42
Valid lab p 11 10 11 11 11 11
Single G cr(1%) 2.564 2.482 2.564 2.564 2.564 2.564
Single G cr(5%) 2.355 2.29 2.355 2.355 2.355 2.355
Single low G 1 0.769 1.121 1.136 0.883 0.900 0.991
Single high G p 2.440 2.243 1.569 2.503 2.243 1.935
Single-low outlier criterion Correct Correct Correct Correct Correct Correct
Single-low outlier Lab
Single-high outlier criterion Straggler Correct Correct Straggler Correct Correct
Single-high outlier Lab Lab 13 Lab 15

Table 5.15 – Results of precision analysis for value Dns [×10-12 m2/s] from Method M6.
Level j PC50 PC42 PC35 SF42 FA42 SL42
Number of valid lab p 10 10 10 10 10 11
General mean m 20.034 19.362 11.12 14.684 6.083 12.198
2
Repeatability variance s r 33.35033 28.48699 3.91065 56.58081 9.51321 21.73269
2
Between-lab variance s L 339.17634 198.61565 70.17476 79.02014 22.62493 125.3162
2
Reproducibility variance s R 372.52667 227.10264 74.08541 135.60095 32.13814 147.04889
Repeatability std. dev. s r 5.775 5.337 1.978 7.522 3.084 4.662
Reproducibility std. dev. s R 19.301 15.07 8.607 11.645 5.669 12.126
Repeatability COV(s r) 28.8% 27.6% 17.8% 51.2% 50.7% 38.2%
Reproducibility COV(s R) 96.3% 77.8% 77.4% 79.3% 93.2% 99.4%
γ = s R/s r 3.34 2.82 4.35 1.55 1.84 2.60
Repeatability limit r = 2.8s r 16.17 14.944 5.538 21.062 8.635 13.054
Reproducibility limit R = 2.8s R 54.04 42.2 24.1 32.61 15.87 33.95
Number of excluded outliers 1 0 1 1 1 0
Outlier type C C C C
Outlier laboratories* Lab 13 Lab 13 Lab 15 Lab 8
* Two abnormal values in Lab 11: PC35-test2 and SL42-test2 have primarily been excluded.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 20 of 38
WP5 REPORT – Final Evaluation of Test Methods

5.4.4 Method R1
For the resistivity measured on the specimens for Method M4 or alike, the results from the
consistency test are shown in Figure 5.6 and Tables 5.16 and 5.17. Three outliers (PC50 from
Lab 11, PC42 from Lab 2 and SL42 from Lab 8) were excluded according to the criteria. The
precision results are shown in Table 5.18.

Within-laboratory Consistency Statistic


According to ISO 5725

PC50
Mandel's k -statistic

PC42
2
PC35
SF42
FA42
SL42
1 1% significance for PC50
5% significance for PC50

0
Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab
1 2 3 4 7 8 9 10 11 12 13 14 15 16 17

Laboratory

Between-laboratory Consistency Statistic


According to ISO 5725

2
PC50
Mandel's h -statistic

PC42
1
PC35
SF42
0
FA42
SL42
-1 1% significance for PC50
5% significance for PC50
-2

-3
Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab
1 2 3 4 7 8 9 10 11 12 13 14 15 16 17

Laboratory

Figure 5.6 - Mandel’s k-statistic and h-statistic of the resistivity (R1 on M4 specimens).
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 21 of 38
WP5 REPORT – Final Evaluation of Test Methods

Table 5.16 - Cochran's test results for the resistivity (R1 on M4 specimens)
Level j PC50 PC42 PC35 SF42 FA42 SL42
Valid lab p 11 12 12 11 11 11
Mean n 3 3 3 3 3 3
C cr(1%) 0.504 0.475 0.475 0.504 0.504 0.504
C cr(5%) 0.417 0.392 0.392 0.417 0.417 0.417
Cochran's test statistic C 0.557 0.284 0.327 0.427 0.457 0.577
Outlier criterion Outlier Correct Correct Straggler Straggler Outlier
Outlier Lab Lab 11 Lab 8 Lab 8 Lab 8

Table 5.17 - Grubbs’ test results for the resistivity (R1 on M4 specimens)
Level j PC50 PC42 PC35 SF42 FA42 SL42
Valid lab p 11 12 12 11 11 11
Single G cr(1%) 2.564 2.636 2.636 2.564 2.564 2.564
Single G cr(5%) 2.355 2.412 2.412 2.355 2.355 2.355
Single low G 1 1.559 0.865 1.186 1.435 1.670 1.885
Single high G p 2.006 2.928 2.088 2.178 1.710 1.628
Single-low outlier criterion Correct Correct Correct Correct Correct Correct
Single-low outlier Lab
Single-high outlier criterion Correct Outlier Correct Correct Correct Correct
Single-high outlier Lab Lab 2

Table 5.18 – Results of precision analysis for the resistivity [Ω⋅m] (R1 on M4 specimens).
Level j PC50 PC42 PC35 SF42 FA42 SL42
Number of valid lab p 10 11 12 11 11 10
General mean m 54.43 60.7 123.8 154.94 322.61 555.43
2
Repeatability variance s r 12.957 41.377 88.19 502.935 1059.372 5792.198
2
Between-lab variance s L 92.566 42.141 597.314 1228.324 10697.766 31921.893
2
Reproducibility variance s R 105.523 83.518 685.504 1731.259 11757.138 37714.091
Repeatability std. dev. s r 3.6 6.43 9.39 22.43 32.55 76.11
Reproducibility std. dev. s R 10.27 9.14 26.18 41.61 108.43 194.2
Repeatability COV(s r) 6.6% 10.6% 7.6% 14.5% 10.1% 13.7%
Reproducibility COV(s R) 18.9% 15.1% 21.1% 26.9% 33.6% 35.0%
γ = s R/s r 2.85 1.42 2.79 1.86 3.33 2.55
Repeatability limit r = 2.8s r 10.08 18 26.29 62.8 91.14 213.11
Reproducibility limit R = 2.8s R 28.8 25.6 73.3 116.5 303.6 543.8
Number of excluded outliers 1 1 0 0 0 1
Outlier type C G C
Outlier laboratories Lab 11 Lab 2 Lab 8
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 22 of 38
WP5 REPORT – Final Evaluation of Test Methods

For the resistivity measured on the specimens for Method M6 or alike, the results from the
consistency test are shown in Figure 5.7 and Tables 5.19 and 5.20. Two outliers (PC42 and
SL42 from Lab 11) were excluded according to the criteria. The precision results are shown
in Table 5.21.

Within-laboratory Consistency Statistic


According to ISO 5725

PC50
Mandel's k -statistic

PC42
2
PC35
SF42
FA42
SL42
1 1% significance for PC50
5% significance for PC50

0
Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab
1 2 3 4 7 8 9 10 11 12 13 14 15 16 17

Laboratory

Between-laboratory Consistency Statistic


According to ISO 5725

2
PC50
Mandel's h -statistic

PC42
1
PC35
SF42
0
FA42
SL42
-1 1% significance for PC50
5% significance for PC50
-2

-3
Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab Lab
1 2 3 4 7 8 9 10 11 12 13 14 15 16 17

Laboratory

Figure 5.7 - Mandel’s k-statistic and h-statistic of the resistivity (R1 on M6 specimens).
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 23 of 38
WP5 REPORT – Final Evaluation of Test Methods

Table 5.19 - Cochran's test results for the resistivity (R1 on M6 specimens)
Level j PC50 PC42 PC35 SF42 FA42 SL42
Valid lab p 7 8 8 8 8 8
Mean n 3 3 3 3 3 3
C cr(1%) 0.664 0.615 0.615 0.615 0.615 0.615
C cr(5%) 0.561 0.516 0.516 0.516 0.516 0.516
Cochran's test statistic C 0.530 0.732 0.423 0.364 0.535 0.637
Outlier criterion Correct Outlier Correct Correct Straggler Outlier
Outlier Lab Lab 11 Lab 14 Lab 11

Table 5.20 - Grubbs’ test results for the resistivity (R1 on M6 specimens)
Level j PC50 PC42 PC35 SF42 FA42 SL42
Valid lab p 7 8 8 8 8 8
Single G cr(1%) 2.139 2.274 2.274 2.274 2.274 2.274
Single G cr(5%) 2.02 2.126 2.126 2.126 2.126 2.126
Single low G 1 0.808 0.796 0.940 1.038 1.340 1.451
Single high G p 1.918 1.931 1.785 2.119 1.689 1.527
Single-low outlier criterion Correct Correct Correct Correct Correct Correct
Single-low outlier Lab
Single-high outlier criterion Correct Correct Correct Correct Correct Correct
Single-high outlier Lab

Table 5.21 – Results of precision analysis for the resistivity [Ω⋅m] (R1 on M6 specimens).
Level j PC50 PC42 PC35 SF42 FA42 SL42
Number of valid lab p 7 7 8 8 8 7
General mean m 105.72 88.38 171.75 206.54 342.67 605.62
2
Repeatability variance s r 620.308 122.136 598.122 1155.182 4407.739 3523.174
2
Between-lab variance s L 4547.924 3155.596 3275.875 5490.262 20158.639 45658.142
2
Reproducibility variance s R 5168.232 3277.732 3873.997 6645.444 24566.378 49181.316
Repeatability std. dev. s r 24.91 11.05 24.46 33.99 66.39 59.36
Reproducibility std. dev. s R 71.89 57.25 62.24 81.52 156.74 221.77
Repeatability COV(s r) 23.6% 12.5% 14.2% 16.5% 19.4% 9.8%
Reproducibility COV(s R) 68.0% 64.8% 36.2% 39.5% 45.7% 36.6%
γ = s R/s r 2.89 5.18 2.54 2.40 2.36 3.74
Repeatability limit r = 2.8s r 69.75 30.94 68.49 95.17 185.89 166.21
Reproducibility limit R = 2.8s R 201.3 160.3 174.3 228.3 438.9 621
Number of excluded outliers 0 0 0 0 0 0
Outlier type
Outlier laboratories
Number of excluded stragglers 0 1 0 0 0 1
Straggler type C C
Straggler laboratories Lab 11 Lab 11
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 24 of 38
WP5 REPORT – Final Evaluation of Test Methods

5.5 Dependency analysis


The weighted regression procedure as specified in ISO 5725 was used for dependency
analysis. Since in most cases the correlation coefficients for three types of relationship are
very similar, as shown in Figure 5.8, the simplest one, that is, s = a⋅m, was used for
describing the dependency. In such a relationship, the slope a describes the mean COV
(Coefficient of Variation), and thus it is very easy to compare the precision of different test
methods. The results are shown in Figures 5.9 to 5.12.

4.5 Actual points


Reproducibility std. dev. sR

4 SR = 0.236m, R² = 0.932

3.5 SR = 0.31 + 0.153m, R² = 0.932


SR = 0.364m^0.733, R² = 0.921
3

2.5

1.5

0.5

0
0 5 10 15 20
-12 2
General mean m (D ns from Method M4, ×10 m /s)

Figure 5.8 – Example of dependency analysis using three different relationships.

5
Actual points of repeatability std dev
4.5
Repeatability/Reproducibility s

Actual points of reprodicibility std dev


4
Sr = 0.152m, R² = 0.857
3.5
SR = 0.236m, R² = 0.932
3

2.5

1.5

0.5

0
0 5 10 15 20
-12 2
General mean m (D ns from Method M4, ×10 m /s)

Figure 5.9 – Dependency of repeatability/reproducibility on mean value (Method M4).


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 25 of 38
WP5 REPORT – Final Evaluation of Test Methods

7
Actual points of repeatability std dev
Repeatability/Reproducibility s

6
Actual points of reprodicibility std dev

Sr = 0.201m, R² = 0.938
5
SR = 0.283m, R² = 0.993
4

0
0 5 10 15 20 25
-12 2
General mean m (D ns from Method D2, ×10 m /s)

12
Actual points of repeatability std dev
Repeatability/Reproducibility s

10 Actual points of reprodicibility std dev


Sr = 0.09m, R² = 0.757
8 SR = 0.135m, R² = 0.88

0
0 10 20 30 40 50 60 70 80
½
General mean m (K from Method D2, mm/yr )

0.3
Actual points of repeatability std dev
Repeatability/Reproducibility s

0.25 Actual points of reprodicibility std dev


Sr = 0.177m, R² = 0.143
0.2 SR = 0.223m, R² = 0.412

0.15

0.1

0.05

0
0 0.2 0.4 0.6 0.8 1 1.2

General mean m (C s from Method D2, mass% of sample)

Figure 5.10 – Dependency of repeatability/reproducibility on mean value (Method D2).


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 26 of 38
WP5 REPORT – Final Evaluation of Test Methods

3
Actual points of repeatability std dev
Repeatability/Reproducibility s

2.5 Actual points of reprodicibility std dev

Sr = 0.22m, R² = 0.918
2
SR = 0.759m, R² = 0.949

1.5

0.5

0
0 1 2 3 4

General mean m (D s from Method M6, ×10-12 m 2/s)

25
Actual points of repeatability std dev
Repeatability/Reproducibility s

Actual points of reprodicibility std dev


20
Sr = 0.357m, R² = 0.388
SR = 0.872m, R² = 0.9
15

10

0
0 5 10 15 20 25

General mean m (D ns from Method M6, ×10-12 m 2/s)

Figure 5.11 – Dependency of repeatability/reproducibility on mean value (Method M6).


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 27 of 38
WP5 REPORT – Final Evaluation of Test Methods

250
Actual points of repeatability std dev
Repeatability/Reproducibility s

Actual points of reprodicibility std dev


200
Sr = 0.105m, R² = 0.968
SR = 0.251m, R² = 0.998
150

100

50

0
0 100 200 300 400 500 600 700

General mean m (ρ from Method R1/M4, Ω ⋅ m)

400
Actual points of repeatability std dev
Repeatability/Reproducibility s

350
Actual points of reprodicibility std dev

300 Sr = 0.16m, R² = 0.705


SR = 0.485m, R² = 0.953
250

200

150

100

50

0
0 100 200 300 400 500 600 700 800

General mean m (ρ from Method R1/M6, Ω ⋅ m)

Figure 5.12 – Dependency of repeatability/reproducibility on mean value (Method R1).


EU-Project CHLORTEST G6RD-CT-2002-00855 Page 28 of 38
WP5 REPORT – Final Evaluation of Test Methods

6 DISCUSSIONS

6.1 Comparison of the precision results


The mean COV-values of repeatability and reproducibility from both WP2 and WP5 are
summarised in Table 6.1 and Figure 6.1. It can be seen that the precision of Method D2
becomes significantly better in WP5, probably due to the fact that the operators in WP5
became better experienced after the practice through WP2. There is no remarkable difference
in precision of Methods M4, M6 (Ds) and R1 (M4) from both WP2 (50-60 mm thick
specimens saturated with demineralised water) and WP5 (50 mm thick specimens saturated
with lime-water), indicating that the defects (air voids or cavities) may not significantly
influence the precision of these test methods. However, there is remarkable increase in COV
for Method M6 (Dns), probably due to the fact that the air voids or cavities can make
chlorides more rapidly penetrating through the specimen. The amount of chlorides
penetrating through the specimen via the air voids or cavities may be limited, depending on
how many defects in the specimen. This amount of chloride may significantly influence the
measurement of time-lag for calculating Dns in Method M6, but may not significantly
influence the total flux of chlorides, resulting in less variation in the measurement based on
the flux of chlorides. The variation in R1 (M6) is larger than that in R1 (M4), probably due to
the fact that the defects more significantly affect the thinner specimens (20 mm in M6 and 50
mm in M4). Therefore, it can be concluded that Method M6 is more sensitive than the other
methods to the defects in specimens.

Table 6.1 – Summary of precision data for various test methods.


Valid number of Repeatability COV (%) Reproducibility COV (%)
Method laboratories*
WP2 WP5 WP2 WP5 WP2 WP5
D2 (Dns) 6∼8 6∼8 29.1 20.1 42.2 28.3
D2 (Cs) 6∼8 6∼8 - 17.7 - 22.3
D2 (K) 6∼8 6∼8 13.2 9.0 20.9 13.5
M4 (Dns) 7∼8 12∼13 17.3 15.2 23.1 23.6
M6 (Dns) 5∼7 10∼11 23.7 35.7 45.4 87.2
M6 (Ds) 5∼7 10∼11 20.8 22.0 79.9 75.9
R1 (M4)** 6∼7 10∼12 9.3 10.5 23.2 25.1
R1 (M6) - 7∼8 - 16.0 - 48.5
* Excluding rejected outliers.
** In WP2, specimens for R1 were 50-60 mm thick and saturated with demineralised water.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 29 of 38
WP5 REPORT – Final Evaluation of Test Methods

50
Repeatability COV, %

WP5
40
WP2
30

20

10

100
Reproducibility COV, %

WP5
80
WP2
60

40

20

0
D2 D2 D2 (K) M4 M6 M6 R1 R1
(Dns) (Cs) (Dns) (Dns) (Ds) (M4) (M6)

Note: In WP2, specimens for R1 were 50-60 mm thick and saturated with demineralised water.

Figure 6.1 – Comparison of COV for various methods.

From Table 6.1 it can be seen that the parameter K from Method D2 reveals the lowest COV
values (repeatability COV 9.0% and reproducibility COV 13.5%). It should be noted,
∂K 1 ∂Dns
however, that K is proportional to the square root of Dns, resulting in ∝ ⋅ , that is,
K 2 Dns
the variation of K is a half of the variation of Dns, as seen in Table 6.1.

Therefore, it can be concluded that, among all the methods for diffusion coefficient (Dns and
Ds), Method M4 reveals the best precision (repeatability COV 15.2% and reproducibility
COV 23.6%).

The resistivity measurement using thick specimens, that is, Method R1 (M4), reveals a better
repeatability COV (10.5%) and a similar reproducibility COV (25.1%) when compared with
Method M4. However, if Method R1 needs to be calibrated against any of the methods for
diffusion coefficient, the variations from both R1 and the calibration method have to be
summed.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 30 of 38
WP5 REPORT – Final Evaluation of Test Methods

6.2 Comparison of the mean values


The mean values of diffusion coefficient and resistivity from both WP2 and WP5 are
summarised in Table 6.2 and Figure 6.2, where the error bars indicate the standard deviation
of reproducibility.

It can be seen that the values of diffusion coefficient measured from both WP2 and WP5 are
in most cases reasonably comparable, especially for the Portland cement concrete with w/c in
the range of 0.42 to 0.50, in which the measured values vary not very much. This is in
agreement with the previous Nordic investigations [2,3], as shown in Figure 6.3. Therefore, it
can be concluded that the qualities of concrete used in WP 5 are in the normal range.

Again, Figure 6.2 shows that the deviation of Dns in Method M6 from WP5 is remarkably
higher than that from WP2, due to the same reasons as discussed in 6.1, that is, the relatively
thin (20 mm thick) specimens used in Method M6, coupled with the presence of defects,
might help to promote the deviation.

Usually it should be easier to saturate a thinner specimen as used in M6 than in M4, resulting
in a lower resistivity. The results in Table 6.1 show, however, that the resistivity measured
with M6 specimen is in general higher than that measured with M4 specimen. Since the
liquid used in M6 (demineralised water) for saturation is different from that used in M4
(saturated lime water) in WP5, the higher resistivity measured by M6 might be attributed to
the less conductive liquid used for saturation, indicating that the liquid used for saturation of
specimens plays important role in this method, especially when specimens contain more
capillary pores, e.g. in Concrete PC50 (w/c 0.50).

Table 6.2 – Summary of mean values of diffusion coefficient and resistivity.


Diffusion coefficient, ×10-12 m2/s Resistivity, Ω⋅m
Concrete Mix D2 (Dns) M4 (Dns) M6 (Dns) M6 (Ds) R1 (M4) R1 (M6)
PC 0.50 17.5 15.4 20.0 2.86 54.4 106
PC 0.45 (WP2) 17.6 16.8 17.8 2.9 70* -
PC 0.42 15.7 14.7 19.4 2.05 60.7 88.4
PC 0.35 4.66 5.94 11.1 1.38 124 172
SF 0.42 4.07 7.44 14.7 1.62 155 207
SF 0.40 (WP2) 2.04 2.61 5.11 1.16 332* -
FA 0.42 1.46 2.31 6.08 0.59 323 343
FA 0.45 (WP2) 7.43 5.4 8.19 1.01 273* -
SL 0.42 1.50 1.77 12.2 0.37 555 606
SL 0.45 (WP2) 2.5 2.51 2.67 0.45 382* -
* In WP2, specimens for R1 were 50-60 mm thick and saturated with demineralised water.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 31 of 38
WP5 REPORT – Final Evaluation of Test Methods

PC50 PC45(WP2) PC42 PC35 SF42 SF40(WP2)

50 15

40

D , x10-12 m2/s
D , x10-12 m 2/s

10
30

20
5
10

0 0
D2 (Dns) M4 (Dns) M6 (Dns) M6 (Ds) D2 (Dns) M4 (Dns) M6 (Dns) M6 (Ds)

FA42 FA45(WP2) SL42 SL45(WP2)

14 10
12
8
D , x10-12 m2/s

10 D , x10-12 m2/s
8 6

6 4
4
2
2
0 0
D2 (Dns) M4 (Dns) M6 (Dns) M6 (Ds) D2 (Dns) M4 (Dns) M6 (Dns) M6 (Ds)

Figure 6.2 – Comparison of the mean values from various methods.

40
D2 in WP2
D ns by D2 and M4, x10-12 m2/s

M4 in WP2
30 D2 in WP5
M4 in WP5
20 D2 in [2]
M4 in [2]
10 D2 in [3]
M4 in [3]

0
0.2 0.3 0.4 0.5 0.6 0.7 0.8
w/c

Figure 6.3 – Comparison of the Dns values by Methods D2 and M4 from various projects.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 32 of 38
WP5 REPORT – Final Evaluation of Test Methods

6.3 Relationships between various transport parameters


Since Method D2 is based on natural immersion under a concentrated condition, which is the
most close to the real situation among all the methods evaluated in WP5, the values from this
method could be used as reference for comparison, as shown in Figure 6.4.

According to Andrade et al [4], resistivity ρ can be converted to a steady state diffusion


coefficient Ds, using the following equation:

k Cl
Ds = (6.1)
ρ

where kCl is a constant. When the dimension of Ds is in cm2/s and ρ in Ω⋅cm, the value of kCl
is of 2×10-4 Ω⋅cm3/s, which was used in WP2 and in this report for comparison.

It can be seen that the diffusion coefficients measured by Methods M4 and D2 are fairly
comparable. This is in agreement with the previous investigations [2,3], as shown in Figure
6.5, where the data from previous investigations are included. Obviously, the most
measurement data are in the range of ± s, where s is the reproducibility standard deviation of
Method D2 (for x-axis) and M4 (for y-axis), respectively. Therefore, Method M4 should be a
good substitution of Method D2 when considering the rapidity, simplicity, cost-effectiveness
and measurement precision.

As expected, the steady state diffusion coefficient Ds (Method M6) is always smaller than the
non-steady state one Dns (Methods D2, M4 or M6). The theoretical explanations have been
given elsewhere [5-7]. The values of Ds in Method M6 from WP2 are better comparable with
Method D2 than from WP5. As discussed in 6.1, the defects, if any, in a specimen or any
leakage from the curved surface of the specimen may cause a shorter time-lag which, in
return, results in a larger diffusion coefficient.

Figure 6.6 shows the relationship between the diffusion coefficients measured by Method M6
and converted from Method R1. It seems the data from both WP2 and WP5 are in fairly good
agreement with the relationship given by Equation (6.1), with the consideration of large
deviation in Method M6. Therefore, it can be concluded that Method R1 should be a good
substitution of Method M6 when considering the rapidity, simplicity, cost-effectiveness and
measurement precision.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 33 of 38
WP5 REPORT – Final Evaluation of Test Methods

25
D from other methods, x10 -12 m 2/s

M4 (Dns)
WP2 data are in solid points
20 M6 (Dns)
M6 (Ds)
15 R1* (M4)
R1* (M6)
10

5 * Converted by
Eq. (6.1) with
k Cl = 200 x10-12
0 3
Ω⋅m /s
0 5 10 15 20 25
D ns from Method D2, x10-12 m 2/s

Note: In WP2, specimens for R1 were 50-60 mm thick and saturated with demineralised water.

Figure 6.4 – Relationships between diffusion coefficients from different methods.

40
D ns from Method M4, x10 -12 m 2/s

WP2
WP5
30
Ref [2]
Ref [3]
20
x±s

10 y±s

0
0 10 20 30 40
D ns from Method D2, x10-12 m2/s

Figure 6.5 – Relationship between diffusion coefficients measured by Methods D2 and M4.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 34 of 38
WP5 REPORT – Final Evaluation of Test Methods

5
R1(M4),
D s from Mthod R1*, x10 -12 m 2/s

WP2
4
R1(M4),
WP5
3
R1(M6),
WP5
2 x±s

1 y±s

0
0 1 2 3 4 5
D s from Method M6, x10-12 m 2/s

Note: In WP2, specimens for R1 were 50-60 mm thick and saturated with demineralised water.

Figure 6.6 – Relationships between diffusion coefficients measured by Method M6 and


converted from Method R1.

7 CONCLUDING REMARKS AND SUGGESTIONS

7.1 Census opinions


Since the opinions to a test method are in many cases very subjective and strongly dependent
on the persons, a statistic way was taken to obtain census opinions to various test methods.
To do so, a questionnaire was prepared and distributed to all partners. Based on the answers
from 10 laboratories to the questionnaire (see Appendix 4), the opinions to the four selected
methods can be summarised in Table 7.1, where the variations (COV%) are given in
parentheses, which reflect the spread of opinions. It can be seen that the opinions to Method
R1 are relatively closer (always lowest variations), while to Method D2 are very discrepant
(always highest variations). According to the census opinions, it can be concluded that the
order of rapidity, simplicity and easiness of the method is R1, M4, M6 and D2.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 35 of 38
WP5 REPORT – Final Evaluation of Test Methods

Table 7.1 – Census opinions to various methods.


Test methods
D2 M4 M6 R1
Rapidity regarding test duration 1.4 (36%) 4.3 (3%) 3.2 (4%) 5 (0%)
Simplicity in equipment
2.4 (22%) 3.7 (9%) 2.9 (12%) 4.4 (6%)
requirement and operation
Easiness in handling specimens 2.2 (27%) 3.6 (9%) 3.7 (7%) 4.3 (5%)
Easiness in measurements 1.7 (28%) 3.8 (6%) 3.5 (6%) 4.4 (7%)
Easiness in calculating results 2.4 (15%) 4 (3%) 2.9 (10%) 4.8 (3%)
Sumary 2 (26%) 3.9 (6%) 3.2 (8%) 4.6 (4%)

7.2 Precision of various methods


From Chapter 6 it can be concluded that the precision data obtained from WP5 are in most
cases comparable with or better than those from WP2, except for Method M6 (Dns), which
was probably affected by some defects in the specimens. Therefore, the data from WP2
should be used for this method. Thus the precision of various methods can be listed in Table
7.2.

Table 7.2 – Precision of methods for testing chloride resistance.


Method Para- Repeatability standard Reproducibility standard
meter deviation sr deviation sR
NT BUILD 443 (D2) Dns sr = 0.201Dns (R2 = 0.94) sR = 0.283Dns (R2 = 0.99)
Cs sr = 0.177Cs (R2 = 0.14) sR = 0.223Cs (R2 = 0.41)
K sr = 0.090K (R2 = 0.76) sR = 0.135K (R2 = 0.88)
NT BUILD 492 (M4) Dns sr = 0.152Dns (R2 = 0.86) sR = 0.236Dns (R2 = 0.93)
Steady state migration Ds sr = 0.220Ds (R2 = 0.92) sR = 0.759Ds (R2 = 0.95)
test (M6) Dns sr = 0.237Dns (R2 = 0.76)* sR = 0.454Dns (R2 = 0.92)*
Resistivity** (R1) ρ sr = 0.105ρ (R2 = 0.97) sR = 0.251ρ (R2 = 998)
* Data from WP2;
** with 50 mm specimen and lime-water saturation.

7.3 Suggestions
Based on the results and experiences from both WP2 and WP5, it can be suggested that

• For the measurement of diffusivity, Method M4 is a good substitution of Method D2


and Method R1 is a good substitution of Method M6, when considering the rapidity,
simplicity, cost-effectiveness and measurement precision;
• Thick specimens, e.g. 50 mm, should be used in Method R1 for resistivity
measurement in order to avoid the influence of defects like air voids or cavities; and
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 36 of 38
WP5 REPORT – Final Evaluation of Test Methods

• Defects like air voids or cavities in the specimen or insufficient sealing of the curved
surface of the specimen may significantly influence the measurement of parameter Dns
when using Method M6.
Although method D2 has the lowest census favourite, it is the only one among all the
evaluated methods, whose exposure condition is the closest to the reality (natural diffusion).
Besides, this method can give some rough information about chloride binding though the
value of Cs. Therefore, it is proposed that the following three methods can be recommended
for further standardisation at the European level:

• Immersion test (based on NT BUILD 443) for determination of non-steady state


diffusion coefficient, Dnssd, and surface total chloride content, Cs;
• Rapid migration test (based on NT BUILD 492) for determination of non-steady state
migration coefficient, Dnssm, under the standardised laboratory exposure condition;
and
• Resistivity test (based on the version used in Appendix 2) for determination of
resistivity ρ as an indirect measurement of the transport property of concrete
All the above three proposed methods have the precision in an acceptable range, that is,
repeatability COV ≤20% (11%∼20%) and reproducibility COV ≤30% (24%∼28%).
Therefore, they are suitable for data exchanges and industrial applications.
Method M6 gives values of both steady state and non-steady state diffusivity, from which
chloride binding can also be estimated. It needs, however, further development for improving
its precision before being standardised as an EN-method for data exchanges and industrial
applications.

REFERENCES

[1] Castellote, M. & Andrade, C., “Pre-evaluation of different test methods”, Chlortest
WP2 report, version 2005-02-25.

[2] Frederiksen, J.M., Sørensen, H.E., Andersen, A. & Klinghoffer, O., ‘HETEK, The
effect of the w/c ratio on chloride transport into concrete - Immersion, migration and
resistivity tests’, HETEK Report, ed. by J.M. Frederiksen, published by the Danish
Road Directorate, 1997.

[3] Tang L. & Sørensen, H.E., “Precision of the Nordic Test Methods for Measuring the
Chloride Diffusion/Migration Coefficients of Concrete”, Materials and Structures, Vol.
34, Oct. 2001, pp. 479-485.

[4] Andrade, C., Alonso, C., Arteaga, A. & Tanner, P., “Methodology based on the
electrical resistivity for the calculation of reinforcement service life”. Supplementary
papers of the proceedings of the Fifth International CANMET/ACI Conference on
Durability of concrete. Barcelona, Spain, 4-9 June 2000, pp 899-915.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 37 of 38
WP5 REPORT – Final Evaluation of Test Methods

[5] Atkinson, A. & Nickerson, A.K., “The diffusion of ions through water-saturated
cement”, J. Materials Sci., 19 (1984) 3068-3078.

[6] Nilsson, L.-O., “A Theoretical Study on the Effect of Non-linear Chloride Binding on
Chloride Diffusion Measurements in Concrete”, Div. of Building Materials, Chalmers
University of Technology, Publication P-92:13, Gothenburg, Sweden, 1992.

[7] Tang, L., “Chloride Transport in Concrete - Measurement and prediction”, Doctoral
thesis, Publication P-96:6, Dept. of Building Materials, Chalmers Universities of
Technology, Gothenburg, Sweden, 1996.
EU-Project CHLORTEST G6RD-CT-2002-00855 Page 38 of 38
WP5 REPORT – Final Evaluation of Test Methods
Appendix 1
1

MULTI-REGIME METHOD: MEASUREMENT OF THE STEADY AND NON-STEADY


STATE CHLORIDE DIFFUSION COEFFICIENTS IN A MIGRATION TEST BY MEANS OF
MONITORING THE CONDUCTIVITY CHANGE IN THE ANOLYTE CHAMBER.

M. Castellote, C. Andrade, and C. Alonso


Institute of Construction Science “Eduardo Torroja”, CSIC, Madrid, Spain.

1. SCOPE

This procedure is for the determination of the steady and non-steady state chloride diffusion
coefficients by monitoring the conductivity of the electrolyte in the anolyte chamber in a
migration experiment.

2. FIELD OF APPLICATION

The method is applicable to hardened specimens cast in the laboratory or drilled from field
structures. The method requires cylindrical specimens of any diameter and a thickness of
15-20 mm, sliced from cast cylinders or drilled cores. The development of this method is
fully described in [1].

3. REFERENCES

[1] M. Castellote, C. Andrade, C. Alonso, Measurement of the steady and non steady state
chloride diffusion coefficients in a migration test by means of monitoring the conductivity
in the anolyte chamber. Comparison with natural diffusion tests, Cement and Concrete
Research, 31 (2001), 1411-1420.

[2] ASTM Standard C 1202–97, Standard Test Method for Electrical Indication of
Concrete’s ability to Resist Chloride Ion Penetration.

[3] C. Andrade, Calculation of chloride diffusion coefficients in concrete from ionic


migration measurements, Cement and Concrete Research, 23, (3) (1993) 724-742.

4. DEFINITIONS

• Migration: Movement of ions under the action of an external electrical field.


• Time-lag : In a diffusion or migration cell, time taken by the chloride ions to
establish a constant flux through the specimen
• Steady-state chloride diffusion coefficient, Ds: Parameter that characterises the
chloride transport in concrete in conditions of constant flux taking into account only
transport phenomena.
• Non steady-state chloride diffusion coefficient, Dns: Parameter that involves the
combined processes of transport and binding of chlorides in concrete.
Appendix 1
2

5. TEST METHOD

5.1 Principle

The test consists of applying a voltage drop to a test specimen between two compartments
containing, upstream, a chloride solution and, downstream, a solution without chlorides.
The negative electrode is placed within the upstream compartment and therefore the
external electrical potential applied forces the chloride ions to migrate through the
specimen towards the downstream compartment.

The test is based in measuring the amount of chlorides arriving in the downstream cell
(anolyte) by means of measuring the conductivity of that solution instead of analysing
chlorides in it. The concentration of chlorides is calculated through a previous empirical
correlation between chloride concentration and conductivity for the specific conditions of
the test.

The steady-state coefficient is calculated then from the flux of chlorides through the
specimen calculated from the measurement of the conductivity of the anolyte in the anodic
compartment.

The calculation of the non steady-state diffusion coefficient is made from the time taken by
the chloride ions to establish a constant flux, that is to say, from the so-called time-lag.

5.2 Reagents and apparatus

5.2.1 Reagents
• Distilled or de-ionised water
• Sodium chloride: NaCl, chemical quality.
• Chemicals for chloride analysis (as required by the analysis method employed)

5.2.2 Apparatus
• Water-cooled diamond saw.
• Vacuum container: capable of containing at least three specimens.
• Vacuum pump: capable of maintaining a pressure of less than less than 1 mm Hg in
the container.
• Migration cell: Different designs can be accepted, provided they include two
different compartments separated by the specimen without any leak. (see figure 1)
• Cathode: black steel grid of the appropriate diameter for the cell or black steel bars.
• anode: black steel grid of the appropriate diameter for the cell or black steel bars.
• Power supply: capable of supplying 12 V DC regulated voltage with an accuracy of
±0.1 V.
• Conductivimeter and conductivity cell
• Voltmeter, capable of displaying voltage ±1 mV.
• Two reference electrodes
• Thermometer or thermocouple with readout device capable of reading to ±1 ºC.
Appendix 1
3

• Slide calliper with a precision of ±0.1 mm.


• Equipment for chloride analysis (as required by the analysis method employed)
• Magnetic stirrer (optional for automatic collection of data)
• Data-logger (optional for automatic collection of data)

5.3 Preparation of the test specimen

5.3.1 Test specimens


If a drilled core is used, the external surface has to be removed in order to avoid the
interference of the chloride ions possibly contained in the surface of the core; only the inner
part of the specimen has to be used. Therefore, if possible, about 30 mm thick layer should
be cut off ( with a water-cooled diamond saw) and the next 15-20 mm thick slice should be
cut as the test specimen.
If a cast cylinder is used, cut the specimen from the central portion of the cylinder as the
test specimen.

5.3.2 Preconditioning (water-saturation)


Prior to the test, the specimens have to be pre-conditioned by saturating it under vacuum in
order to avoid the transport of chlorides by capillary absorption. The procedure adopted in
this method is those given in [2], which can be summarised as follows:

• Place the test specimens in a container (previously rinsed with demineralized water)
or vacuum desiccator. Both end faces of specimen must be exposed.
• Seal desiccator and start vacuum pump. Pressure should decrease to less than 1 mm
Hg (133 Pa) within a few minutes.
• Maintain vacuum for 3 h.
• Fill separatory funnel or other container with de-aerated water (vigorously boiled
tapwater).
• With vacuum pump still running, open water stopcock and drain sufficient water
into the container to cover the specimens (do not allow air to enter desiccator
through this stopcock).
• Close water stopcock and allow vacuum pump to run for one additional hour.
• Close vacuum line stopcock, then turn off pump.
• Turn vacuum line stopcock to allow air to re-enter desiccator.
• Keep the specimens under water (the water used in the previous steps) for 18 ± 2 h.

5.4 Procedure

5.4.1 Measurement of dimensions

• Measure the diameter of the test specimens at two positions at right angles. The
mean of the measurements is taken as the diameter D. (see fig. 1-a)
Appendix 1
4

• If the cell device implies the use of some of these area for sealing the cell, only the
diameter of free path for the ions has to be taken into account. (see fig. 1-b)

• Measure the thickness of the test specimen at 4 positions equally spaced around the
circumference. The mean of the 4 measurements is taken as the thickness, l. (see fig.
1-c)

a) b) c)
Figure 1: Schematic representation of the measurement of dimensions of the specimen

5.4.2. Assemblage of the cell

Fit the specimen in the cell, by the appropriate mechanism depending on the type of cell,
introducing the electrodes into the corresponding holes (depending if using bars or plates
for the electrodes, the electrodes have to be mounted from the beginning or after having the
cell assembled). (see figure 2)

Fill in the downstream compartment with distilled or de-ionised water. Wait for a few
minutes and check that there are no leaks. Then fill in the upstream solution (1 M NaCl
solution). The exact amount introduced in both compartments, Va, Vc (ml) has to be
recorded precisely and the exact concentration of the chloride solution has to be known by
analysis.

Data can be taken automatically by means of a conductimeter and voltmeter connected to a


data-logger. If so, introduce the conductivity cell into the anodic compartment and the
reference electrodes in both compartments as close as possible from the sample. If the
conductivity cell is not equipped with a temperature sensor and correction of temperature,
introduce also a termocouple and connect it to the data-logger. It is very important to
calibrate the conductivity cell, with the appropriate standard solution, prior and at the end
of the test.

Data can be also taken manually; if so, the conductivity cell and reference electrodes have
to be introduced in the corresponding compartments only in the moment of taking the data,
and they are kept outside until next measurement. In this case, it is recommended to
calibrate the conductivity cell before each measurement.
Appendix 1
5

Connect the cathode to the negative pole and the anode to the positive pole of the power
supply.

- Power supply
12 V DC
+

-
∆V
+
Black steel
electrodes Temp. meas.

NaCl DOWNSTREAM
Distilled water (ANOLYTE)
UPSTREAM 1M
(CATHOLYTE)
Conductivity cell
reference electrodes
Magnetic stirrer
specimen

Figure 2: Schematic representation of the set up of the test for automatic data collection.

5.4.3. Migration test

Turn on the power supply. Set the voltage at 12 Volts ± 0.1 V.

At periodic intervals, the effective voltage, the conductivity and temperature in the anolyte
compartment have to be measured.

For doing that, as a first step, the effective voltage drop has to be taken by means of placing
two reference electrodes near both faces of the specimen (see detail in figure 3) and then,
the current has to be switched off during 5 minutes during which the anolyte liquid is
stirred by the magnetic stirrer. After these minutes, the stirrer is switched off and the
Appendix 1
6

measurement of the temperature and conductivity of the anolyte can be taken. Then,
connect again the power supply.

- Power supply
12 V DC
+

-
∆V
+

reference electrodes

Figure 3: Schematic representation of the set up of the test: taking the data of effective voltage
drop

If the conductimeter is equipped with automatic correction of the values to the reference
temperature of 25ºC, measurement of the temperature is not necessary.

If data are taken automatically the frequency of measurements can vary depending on the
equipment; as a general rule a data every 3-6 hours can be enough. If data are taken
manually, at least a recording of the corresponding data every 12 hour has to be taken and a
magnetic stirrer is not necessary as the liquid is homogenized when introducing and taking out
the different electrodes and conductivity cell.

It has to be pointed out that the oxidation of the anodic electrode (ordinary steel) takes
place and therefore the anodic solution becomes brownish. This is necessary for avoiding
evolution of chlorine gas at the anode. These oxides do not perturb neither the experiment
nor the measurement of conductivity.

The evolution of the conductivity in the anodic chamber follows the trend in three steps
given in figure 4. The test is finished when step 3 has been clearly reached, and depends on
the type of sample. As a general rule, two weeks are enough for most of samples.
Appendix 1
7

mmol Cl- anolyte


60

50

40
steady-state
30 Step Step
20 1 3
Step
10
2
0
0 50 100 150 200 250 300 350
τ time

Figure 4: Schematic representation of the evolution of the conductivity and amount of


chlorides in the anolyte during the test.

6. TEST RESULTS AND CALCULATIONS

To make this part easy to understand, an example is given simultaneously with the
description of the steps. Data for the example are given in table 1.

For the calculation of the steady state and non steady state diffusion coefficients by this
method, the following steps have to be followed:

1: Record the evolution during the test of the following parameters: effective voltage Ve
(V), conductivity (mS/cm) and temperature (ºC) (if no automatic correction for
conductivity is provided by the conductimeter).

2: Calculate the values of conductivity at the reference temperature of 25ºC using equation
(1)
c 25 =c T +0.02(25 − T )cT (1)
Where:
c25 = conductivity (mS/cm) at 25ºC.
cT = conductivity (mS/cm) measured at temperature T (ºC).

3: Calculate the values of accumulated chlorides according to equation (2) [1]

Cl-(mmol) = (-1.71+ 11.45 * c25 ) Va. (2)


Appendix 1
8

where Va is the volume of the anolyte (l)

If according to the equation (2) [Cl-] (mmol) <0, then take Cl-(mmol) = 0

Voltage applied (V) 12 l (cm) 1


initial [Cl] catholyte (mol/l) 1.011 φ (cm) 7.5 Va (l) 0.414
time effective voltage conductivity, cT Temperature corrected conductivity, c25 Cl anolyte
(hours) (V) mS/cm (ºC) mS/cm mmol
0.00 3.20 0.09 19.50 0.10 0.00
3.00 4.00 0.17 19.50 0.19 0.19
16.67 6.77 0.50 18.50 0.57 1.97
24.50 5.14 0.33 18.50 0.37 1.06
40.75 6.33 0.36 18.00 0.41 1.24
47.92 6.70 0.44 18.00 0.50 1.67
66.09 8.27 0.97 17.00 1.13 4.62
72.25 7.92 1.39 16.50 1.63 7.00
88.00 8.65 2.74 15.50 3.26 14.74
97.75 9.46 3.89 17.00 4.51 20.68
115.42 9.86 5.30 14.00 6.47 29.94
139.75 10.30 6.54 14.00 7.98 37.11
150.00 10.50 6.52 15.00 7.82 36.37
Table 1: Example of the test results and calculations. Steps 1-3.

4: Draw the evolution of the amount of chlorides in function of the time and from it, select
the period of the steady state and therefore the points indicating the steady state initiation (ssi)
and the end of the steady state (ssf). During this period, the corresponding points have to be
fitted to a linear regression (see figure 5). In the case of the example, this period can be taken
from 66.09 (ssi) to 115.42 hours (ssf), where the correlation obtained is very good.

40 60

35 50 y = 0.5196x - 30.28
R2 = 0.9977
40
30
ssf 30
mmol Cl anolyte

25
mmol Cl anolyte

20
20
10
15 0
0 50 100 150
10 -10

5 -20
ssi
-30
0
0 50 100 150 -40
time (hours) time (hours)

Figure 5: Example of the test results and calculations. Step 4.


Appendix 1
9

5: Calculate the steady-state diffusion coefficient, Ds, from the Modified Nernst-Planck
equation (eq.3) [3].

J Cl RTl
Ds = (3)
z F C1 γ∆Φ
where:

(mmol ssf − mmol ssi )1E − 3


J =
St ss ( f −i )

J = Flux of chlorides through the specimen during the steady-state period (mol/cm2s)
S : surface area of test specimen exposed to the chloride solution (cm²)
t: time (in seconds) of duration of the select the period of the steady state.

From the data of the example:

J = 3.226E-9 mol/cm2s
C1= Cl- concentration in the catholyte (mol/cm3)= 1.011E-3
γ = Activity coefficient of the catholyte solution= 0.657
∆Φ= Averaged effective voltage (V) through the specimen during the steady-state period= 8.9
l = thickness of the sample (cm) =1.
R : perfect gas constant (1.9872 cal/mol K)
T : Average temperature during the test (K) =289.55
z: ion valence, for chloride, z = 1;
F : Faradays constant = 23060 (cal/ V eq)

According to these data, Ds = 1.36E-8 cm2/s

6: Calculate the non steady-state diffusion coefficient, Dns, from the time-lag, t, obtained by
the intersection of the straight line of chloride flux characteristic of the steady state with the X
axis, and later application of equation (4).

2l 2  v 
Dns = 2 v coth − 2 (4)
τv  2 
where:

τ = time-lag in the migration test (s); It is obtained from the linear correlation obtained in
step 4, calculating the time for chloride in anolyte equal to zero.
Appendix 1
10

In our example, τ = 58.3 hours = 209880 s

l = thickness of the specimen (cm)


v= ze(∆φ)t/kT; where

k= Boltzmann´s constant.
T : Average temperature during the test (K) =289.55

(∆φ)t: Averaged effective voltage (V) through the specimen from the beginning of the test
until the time lag = 5.92

v= ze(∆φ)t/kT= 1.16E4 x 5.92 /289.55 = 237.26

For the values of voltage drop usually used in the test, equation (4) can be simplified into
equation (5)
2l 2 (v − 2)
Dns = (5)
τ v2
Therefore

Dns = 3.98E-8 cm2/s

7. TEST REPORT

The test report should include the following information:


• Name and address of the test laboratory.
• Date of arrival of the material or object tested.
• Brief description of the material or object tested, including if it is a core of a cast
specimen, composition, curing age and date of arrival.
• The name of the method.
• Name and address of the person who performed the test.
• Date of the test.
• Test results, including all the items given in table 2 and the obtained values for the
averaged effective voltage during the steady-state period, averaged effective voltage
from the beginning of the test until the time lag, and the time-lag obtained in the
migration test.
• Calculated Ds and Dns
• Any observation of incidences during the test or any deviation from the test method.
• Date and signature.
Appendix 2

DETERMINATION OF ELECTRICAL RESISTIVITY IN CONCRETE


SPECIMENS.

DIRECT METHOD.

(Translation of UNE 83XXX) made by C. Andrade

1. Scope and field of application


This standard describes the method for the determination of electrical resistivity of
hardened concrete in specimens or cores.

The testing samples will be concrete specimens or cores not containing reinforcements
and of regular geometry. The specimens will be of the shape and size described in the
Standards UNE83-301: 1991.

Concrete resistivity of a water saturated concrete is an indirect measurement of


concrete pore connectivity. In a non saturated concrete it is an indication of its degree
of satuaration.

2. Standards for consult


UNE-83301: 1991 – Ensayos de hormigón. Fabricación y conservación de probetas.
UNE-83302: 1984 – Ensayos de hormigón. Fabricación y conservación de testigos.

3. General requirements for the testing


3.1. Number of determinations

The number of determinations will be of two in each specimen or core.

3.2. Expression of electrical resistance, resistivity, dimensions and results

The electrical resistances will be expressed in Ohms, the dimensions in meters, the
areas in square meters and the resistivities in ohms per meter.

The results of resistivity of tests, given as the average of the two determinations, will be
expressed in ohm·meter in full numbers.

3.3. Conditions of testing room

The temperature of the testing room has to be between 20±2ºC and the relative
humidity, HR, cannot be smaller than 45%.

3.4. Conditions for maintenance of testing specimens or cores

The samples to be tested will be cured or maintained as indicated in paragraph 7 of


this standard. In any case the specimens or cores cannot remain out of the saturation
Appendix 2

in water (humid chamber or under water) more than 5 minutes. In this period of time
the measurements have to be carried out.

4. Fundamentals of the method


It is based in the measurement of the electrical resistance of a concrete sample by
means of current lines parallel to the basement of the sample. This method is
considered as the reference one for the measurement of resistivity in concrete
specimens or cores.

5. Definitions
5.1. Electrical resistance of concrete, Re

It is the relation between the voltage drop (V) between the internal electrodes and the
current (I) that circulates between the two external ones through the specimen or core.

5.2. Electrical resistivity of concrete, ρ

It is the electrical resistance by unit of volume of concrete. It is the inverse of the


conductivity and it is obtained from the relation between the voltage drop and the
current refereed to a standarized geometry (a cube of 1m of edge).

The resistivity of a concrete saturated in water is an indirect measurement of the


concrete pore connectivity.

In a non water-saturated concrete represents an indication of the degree of saturation.

5.3. Cell constant (k)

It is the relation between the electrical resistance and the resistivity. It is obtained
through Ohm’s law.

NOTE: The quantitative parameters of all definitions are not object of conformity.

5.4. Vacuum saturation

It is such obtained through immersion under vacuum in the conditions specified in


paragraph 7 o this standard.

6. Apparatus
6.1. Resistivimeter

It is an equipment able to apply a stable voltage or current to the test specimen and to
measure respectively the current or voltage drop generated. It is recommended that the
equipment use an alternating signal in such a way that in the circuit circulates a current
of about 40mA at a frequency between 70-100Hz. It will be also acceptable to use an
alternating current source with two external multimetres to measure the current and
voltage drop. A datalogger for monitoring and simultaneous measurement is also
allowed.
Appendix 2

6.2. Electrodes

Metallic nets (openings < 2mm) or plates will be used as electrodes. They can be of
steel, copper or any other well conductor metal, free of superficial impurities (deposits,
rust, oxides, etc.) and of equal dimensions of that of the transversal section of the
specimen or core (see figure 1).

6.3. Contact sponges

Two thin sponges (thickness < 5mm) of equal size than the electrodes, that will be
placed between them and the two parallel faces of the specimen.

6.4. Water for wetting the sponges

The water for wetting the sponges will be tap water.

6.5. A mass of about 2 kg

Any object with at least a place face with similar dimensions than the sample face,
made of 9 non conductor material, that will be used to press the upper electrode
against the contact sponge and the concrete (see figure 1).

7. Sample preparation
The test must be performed in a fully water saturated specimen. These conditions are
fulfilled in specimens cured submerged in water. In this case care should be taken to
avoid any superficial evaporation from the top surface of the specimen durring the first
24 hours when the specimens remain inside the moulds. To avoid this evaporation
plastic covers are suitable.

For the rest of the cases, water saturation will be achieved through the following
procedure: the samples will be introduced in a hermetic container (for instance a
desiccator) with two holes in order to allow making vacuum and introducing the water.
After the sealing f the container, vacuum will be made to reduce the internal pressure
to less than 1 mm Hg. This vacuum level will be maintained during 3 hours. With the
vacuum pump functioning, the decarbonated water will be introduced in such a quantity
to completely cover the samples. The vacuum will be maintained during 1 further hour.
After switching off the vacuum pump, air will be allowed to enter in the container. The
samples will be immersed during 18±2 additional hours. Then, the specimens will be
maintained immersed or in a chamber at RH>95% until the resistivity determination.
samples will be immersed during 18±2 additional hours. Then, the specimens will be
maintained immersed or in a chamber at RH>95% until the resistivity determination.

8. Testing method
The steps of the test will be the following:

8.1. Connections arrangement


Appendix 2

The connection arrangement is that of figure 1. The two electrodes are plugged to the
resistivimeter.

Mass of
2Kg

ELECTROD Resistivimeter
sponges

SPECIMEN

HOLDE

Figure 1. Arrangement for the measurement of resistivity by the direct method.

8.2. Wetting and measurement of resistance of the sponges

The sponges will be wetted with tap water complying with the standard UNE 83301.
They will be slightly dripped in order to remove the excess of water. They will be placed
between the two electrodes and the mass of 2 kg will be placed on the top of the upper
electrode. Then, the current will be applied with the resistivimeter and the potential
drop measured as soon as the measurement is stable (after around 5 seconds) in
order to obtain the electrical resistance of the sponges, Rsp, alone. This Rsp will not be
higher than 100 Ω.

8.3. Measurement of the electrical resistance

Just before the measurement, the dipping water will be removed from the surface of
the samples, in particular the lateral surfaces, with a cotton cloth. The sample will be
placed vertically lying on one electrode and a contact sponge as shown in figure 1. On
the top, the other sponge and electrode will be placed and on the top of them, the mass
of 2Kg. Then, the measurement will be carried out and the electrical resistance of the
sample plus the sponges (Re+sp) will be recorded after stabilization of the signal.

9. Calculation and expression of results


9.1. Calculation of the electrical resistance of the sample

The electrical resistance will be obtained from the expression:


Appendix 2

Re= Re+sp - Rsp

where Re is the electrical resistance of the sample


Re+sp is the electrical resistance of the sample plus the sponges
Rsp is the electrical resistance of the sponges

9.2. Calculation of cell constant

The cell constant for cylindrical shape in specimens of 15cm in diameter and 30cm in
height will be calculated as follows:

S
K=
L

where S the area of the sample face where the sponge is placed in m2
L is the height of the specimen or core in m

9.3. Calculation of the resistiviy

The resistivity is then calculated from

ρ = k Re

where ρ is the resistivity in Ω·m


k is the cell constant in m
Re is the electrical resistance in Ω

9.4. Expression of result

The resistivity will be given in Ω·m in full numbers and will be calculated for each
sample as the average of the six measurements carried out.

It will be indicated the temperature in ºC at which the measurement was made.

10. Precision and bias


No reproducibility and repetitivity indications are available.

11. References
APM –219 : 1996. Laboratory AEC (Denmark)

ASTM – G 57 – 78 (modification 1984).

RILEM Recommendation.- Test methods for on-site measurement of resistivity of


concrete.- Materials &Structures vol 33, December 2000, 603-611.
Appendix 2

Annex 1 (Informative)
This standard can be applied to concrete for public works and buildings made with
standard aggregates (for instance: siliceous, limestone, etc.). It cannot be applied
therefore to concretes made with highweight, heavy or conductor aggregates, as well
as concretes reinforced with metallic fibers.
Appendix 3
1 (8)
Appendix 3: Summary of the results from individual laboratories
-12 2
Collection of the Original Data y ij Method: D2 x10 m /s
Decimal: 1 1 2 2 2 2
Laboratory PC50 PC42 PC35 SF42 FA42 SL42
Lab 1, test 1 19.3 19.8 6.31 5.37 1.72 1.87
test 2 26.5 17.2 5.74 4.75 1.3 1.46
test 3 19.9 18.8 5 4.03 1.86 1.06
Lab 2, test 1 15.9 16.2 4.06 4.57 1.92 0.93
test 2 1.35
test 3
Lab 3, test 1 12.7 7.7 2.91 2.61 1.19 0.91
test 2 13.8 10.3 1.91 2.66 1.29 1.13
test 3 13.1 10.3 3.57 2.83 0.93 1.12
Lab 4, test 1 19 28.8 5.31 4.07 1.51 1.74
test 2 28.6 19.4 5.07 5.56 1.57 1.82
test 3 22.8 18.5 5.94 5.19 1.59 1.82
Lab 7, test 1
test 2
test 3
Lab 8, test 1
test 2
test 3
Lab 9, test 1 14.5 4.9 1.58 2.24
test 2 13.6 4.56 1.41 1.32
test 3 15.8 1.62 2.49
Lab 10, test 1
test 2
test 3
Lab 11, test 1 14.7 28.8 4.5 5.44 1.13 1.49
test 2 16.7 11.6 5.22 4.91 1.6 1.27
test 3 18.3 12.2 5.61 4.28 1.6 1.17
Lab 12, test 1 13.6 14.8 4.03 2.2 1.24 1.19
test 2 12.8 13.7 4.07 2.33 1.28 1.24
test 3 13 10.8 3.38 2.59 0.87 1.26
Lab 13, test 1 17.3 7.1 5.47 4.44 0.32 1.6
test 2 15.6 13 9.07 3.9 1.94 0.82
test 3 31 22 4.83 3.91 1.55 2.19
Lab 14, test 1 12.9 16.1 5.22 4.27 1.86 1.46
test 2 14.6 15.4 5.52 4.35 1.54 1.2
test 3 12.1 15.4 4.57 4.52 1.5 2.09
Lab 15, test 1
test 2
test 3
Lab 16, test 1 56.3 29.2 16.94 52.35 11.94 6.83
test 2 57.9 45.6 27.06 55.75 19.86 5.61
test 3 67.3 30.2 36.81 64.65 12.92 4.5
Lab 17, test 1
test 2
test 3
Appendix 3
2 (8)

Collection of the Original Data y ij Method: D2 (Cs) mass%


Decimal: 3 3 3 3 3 3
Laboratory PC50 PC42 PC35 SF42 FA42 SL42
Lab 1, test 1 0.895 0.792 0.599 0.641 1.095 0.625
test 2 0.671 0.763 0.648 0.619 1.209 0.715
test 3 0.768 0.689 0.908 0.847 0.983 1.046
Lab 2, test 1 0.673 0.622 0.626 0.68 0.674 0.673
test 2 0.667
test 3
Lab 3, test 1 0.782 0.839 0.814 0.902 1.106 0.747
test 2 0.743 0.814 0.851 0.972 0.99 0.767
test 3 0.741 0.75 0.715 0.877 1.427 0.625
Lab 4, test 1 0.806 0.65 0.646 0.612 0.978 0.656
test 2 0.726 0.752 0.659 0.589 1.077 0.579
test 3 0.745 0.718 0.594 0.665 0.832 0.615
Lab 7, test 1
test 2
test 3
Lab 8, test 1
test 2
test 3
Lab 9, test 1 0.612 0.902 1.386 0.608
test 2 0.634 1.033 1.183 0.799
test 3 0.613 1.149 0.62
Lab 10, test 1
test 2
test 3
Lab 11, test 1 0.773 0.663 0.711 0.804 1.062 0.605
test 2 0.695 0.87 0.634 0.856 0.865 0.67
test 3 0.684 0.708 0.678 0.86 0.983 0.748
Lab 12, test 1 0.991 0.909 0.946 1.347 0.699 0.664
test 2 0.997 0.935 1.023 1.169 1.177 0.356
test 3 1.084 0.861 1.328 0.911 1.688 0.706
Lab 13, test 1 0.864 1.204 0.546 0.796 1.842 0.572
test 2 0.565 0.607 0.464 0.921 1.12 1.178
test 3 0.507 0.513 0.534 0.857 1.077 0.285
Lab 14, test 1 0.794 0.824 0.76 0.847 0.982 0.879
test 2 0.725 0.773 0.794 0.917 1.039 1.046
test 3 0.719 0.781 0.793 0.88 1.173 0.799
Lab 15, test 1
test 2
test 3
Lab 16, test 1 0.71 0.773 0.747 0.577 0.694 0.256
test 2 0.626 0.599 0.647 0.574 0.524 0.339
test 3 0.714 0.757 0.575 0.516 0.668 0.436
Lab 17, test 1
test 2
test 3
Appendix 3
3 (8)

1/2
Collection of the Original Data y ij Method: D2 (K) mm/year
Decimal: 1 1 1 1 1 1
Laboratory PC50 PC42 PC35 SF42 FA42 SL42
Lab 1, test 1 72 71.1 36.4 34.7 22.6 20.8
test 2 79.1 65.6 35.4 32.4 20 19
test 3 70.7 67.2 35.6 32 22.9 17.5
Lab 2, test 1 55 55.7 27.9 29.5 19.1 13.3
test 2 16
test 3
Lab 3, test 1 56 47 37 35 23 20
test 2 59 54 29 34 24 24
test 3 62 52 40 38 22 19
Lab 4, test 1 69.4 81.4 34.4 29.7 20.8 19.8
test 2 83.2 69.1 33.7 34.4 21.6 19.8
test 3 74.8 66.7 35.6 34.2 20.6 20
Lab 7, test 1
test 2
test 3
Lab 8, test 1
test 2
test 3
Lab 9, test 1 56 36 32 23
test 2 55 36 22 19
test 3 59 23 24
Lab 10, test 1
test 2
test 3
Lab 11, test 1 56 78 31 34 17 17
test 2 58 52 32 33 19 16
test 3 61 51 34 31 20 16
Lab 12, test 1 58.7 60.3 31.7 25.4 16.8 16.1
test 2 57 58.4 32.4 25.5 19 13.8
test 3 58.6 51 31.2 25.4 16.8 16.8
Lab 13, test 1 63 43.3 31.8 31.3 10 17.4
test 2 54.1 50.2 38.9 30.4 22.3 14.7
test 3 73.7 61.9 29.6 29.9 19.8 16.1
Lab 14, test 1 54 61 35 32 23 19
test 2 56 59 36 32 20 18
test 3 51 59 33 32 20 23
Lab 15, test 1
test 2
test 3
Lab 16, test 1 113 82.9 61.2 101.9 52.3 29
test 2 111.4 97.9 74.9 105.1 63.3 29.1
test 3 123.7 84 85 109.9 54 28.2
Lab 17, test 1
test 2
test 3
Appendix 3
4 (8)

-12 2
Collection of the Original Data y ij Method: M4 x10 m /s
Decimal: 1 1 2 2 2 2
Laboratory PC50 PC42 PC35 SF42 FA42 SL42
Lab 1, test 1 15.5 13.9 5.53 7.51 2.19 2.37
test 2 15.1 20.9 5.46 5.26 2.36 2.28
test 3 15 19.7 5.14 7.97 2.08 2.19
Lab 2, test 1 14.7 14.9 7.29 3.25 1.88
test 2 16.5 13.3 6.08 5.76 3.09 1.57
test 3 17.1 11.5 7.1 5.24 3.82 1.9
Lab 3, test 1 14.2 15.1 4.51 7.59 2.08 2.04
test 2 17.1 14.4 3.79 9.63 2.55 1.84
test 3 13.7 13 6.31 8.73 2.62 1.39
Lab 4, test 1 19.3 20.7 5.1 8.2 1.62 1.61
test 2 16.8 17.5 5.54 9 1.36 2.37
test 3 15.9 17 5.22 6.92 2.22 1.36
Lab 7, test 1 11.1 11 5.82 7.07 2.83 1.5
test 2 11.6 9.8 7.64 6.56 3.12 1.84
test 3 11.8 10 8.71 6.06 3.63 1.56
Lab 8, test 1 10.1 11.9 5.79 5.65 1.48 1.21
test 2 11.1 12.8 6.31 6.5 2.85 1.19
test 3 11.7 11.5 5.46 6.5 2.43 1.78
Lab 9, test 1 13.4 12.9 5.85 8.71 3.43 1.61
test 2 17 11.6 6.03 6.51 1.95 1.22
test 3 17.3 9.8 6.9 6.6 2.21 1.16
Lab 10, test 1
test 2
test 3
Lab 11, test 1 17.9 14.9 5.37 7.39 1.8 1.05
test 2 16.1 17.1 7.47 6.78 1.6 1
test 3 16.2 20.1 5.26 6.39 1.56 0.94
Lab 12, test 1 16.2 14.3 4.92 5.21 1.4 1.5
test 2 14.3 16.5 5.6 4.6 1.41 1.81
test 3 15.8 15.7 4.3 5.19 3.17 1.43
Lab 13, test 1 21.3 15.7 5.56 8.56 1.54 0.78
test 2 17.7 17.4 5.78 9.8 1.25 1.19
test 3 18.1 5.34 10.37 1.38 2.39
Lab 14, test 1 13.2 12.8 6.99 7.13 2.44 2.71
test 2 15.5 12.6 6.8 7.01 2.25 1.98
test 3 14.6 12.9 6.98 7.41 2.65 2.51
Lab 15, test 1 17.7 18 6.05 8.19 2.06 1.48
test 2 18.3 13.1 6.2 9.15 5.95 1.6
test 3 16.7 19.1 6.81 10.61 2.23 2.02
Lab 16, test 1
test 2
test 3
Lab 17, test 1 5.38 8.16 1.9 2.73
test 2 5.87 8.69 2.94 3.54
test 3 5.54 10.16 2.81 2.29
Appendix 3
5 (8)

-12 2
Collection of the Original Data y ij Method: M6 (Dns) x10 m /s
Decimal: 2 2 2 2 2 2
Laboratory PC50 PC42 PC35 SF42 FA42 SL42
Lab 1, test 1 7.99 9.14 4.84 12.7 8.55 7.49
test 2 7.46 10.7 4.34 10.31 17.26 12.45
test 3 6.53 16.61 6.99 3.83 3.67
Lab 2, test 1 18.06 11.89 9.04 17.83 3.5 3.7
test 2 13.85 14.73 7.22 7.57 4.73 3.94
test 3 17.38 16.67 7.5 13.36 4.27 3.86
Lab 3, test 1 2.76 3.97 1.8 3.97 1.05 1.03
test 2 1.65 2 1.84 6.35 0.74 1
test 3 6.91 2.68 2.35 1.77 1.53 1.09
Lab 4, test 1
test 2
test 3
Lab 7, test 1
test 2
test 3
Lab 8, test 1 21.63 16.53 13.14 41.41 38.57 31.18
test 2 20.43 14.09 12.44 17.69 30.06 33.47
test 3 21.44 15.59 14.34 13.39 10.7 18.38
Lab 9, test 1 13.13 18.28 6.18 7.51 3.96 1.65
test 2 13.04 17.09 5.59 5.63 3.27 1.96
test 3 12.25 15.58 6.57 7.21 2.61 1.73
Lab 10, test 1
test 2
test 3
Lab 11, test 1 19.36 19.03 24.07 14.08 3.33 20.57
test 2 31.48 19.28 105.73 43.01 5.51 741.44
test 3 19.57 20.74 19.8 30.58 4.38 9.88
Lab 12, test 1 9.06 7.64 2.16 3.32 1.26 1.11
test 2 10.44 9 2.4 3.93 1.56 1.2
test 3 13.2 14.21 4.09 5.08 1.42 1.64
Lab 13, test 1 158.91 47 21.13 28.32 5.8 6.4
test 2 71.15 65.02 22.31 26.15 5.43 5.64
test 3 39.34 30.37 6.8 5.57
Lab 14, test 1 10.63 15.75 12.44 4.72 5.01 17.6
test 2 11.77 13.71 7.4 10.9 3.66 13.98
test 3 18.7 22.17 4.82 6.41 6.6 19.33
Lab 15, test 1 92.2 23.82 34.13
test 2 26.15 82.01 14.14 25.52
test 3 64.91 59.29 27.92 40.04 26.37 43.13
Lab 16, test 1 37.43 39.02 20.48 27.48 8.97 25.45
test 2 24.33 30.33 22.05 10.86 9.45 15.39
test 3 33.39 53.05 16.56 21.63 11.42 17.19
Lab 17, test 1
test 2
test 3
Appendix 3
6 (8)

Collection of the Original Data y ij Method: R1 (M4) Ω⋅m


Decimal: 1 1 1 1 1 1
Laboratory PC50 PC42 PC35 SF42 FA42 SL42
Lab 1, test 1 59 51 142 167 495 555
test 2 55 51 137 179 435 569
test 3 56 47 148 150 490 560
Lab 2, test 1 129 157 210 502
test 2 95
test 3 74 110 91
Lab 3, test 1 52 54 149 199 331 611
test 2 58 56 149 208 344 593
test 3 56 70 149 192 318 646
Lab 4, test 1 71 66 144 153 512 684
test 2 62 65 152 124 508 618
test 3 63 73 144 194 487 713
Lab 7, test 1 54 57 101 125 207 381
test 2 54 55 96 117 209 386
test 3 51 53 102 119 214 379
Lab 8, test 1 72 56 117 97 368 219
test 2 71 69 121 184 295 669
test 3 77 55 92 178 441 735
Lab 9, test 1 45 61 122 156 354 613
test 2 51 65 121 159 315 682
test 3 46 76 115 138 356 686
Lab 10, test 1
test 2
test 3
Lab 11, test 1 70 64 171 267 313 1019
test 2 53 74 162 205 368 783
test 3 45 67 194 237 290 713
Lab 12, test 1 42 62 118 164 340 716
test 2 45 53 120 171 352 570
test 3 43 56 127 136 330 738
Lab 13, test 1
test 2
test 3
Lab 14, test 1 61 71 110 137 301 222
test 2 55 70 96 125 209 484
test 3 54 73 106 119 282 342
Lab 15, test 1 39 50 116 143 250 556
test 2 40 51 112 131 253 613
test 3 41 48 117 135 238 556
Lab 16, test 1 61 76 108 96 155 215
test 2 52 58 78 112 154 256
test 3 47 50 111 96 132 204
Lab 17, test 1
test 2
test 3
Appendix 3
7 (8)

Collection of the Original Data y ij Method: R1 (M6) Ω⋅m


Decimal: 1 1 1 1 1 1
Laboratory PC50 PC42 PC35 SF42 FA42 SL42
Lab 1, test 1 129 95 194 195 516 799
test 2 101 77 204 210 344 967
test 3 87 89 139 186 350 981
Lab 2, test 1 69 61 116 145 195 274
test 2 74 76 124 151 201 318
test 3 67 67 109 143 186 283
Lab 3, test 1 58 70 90 101 165 493
test 2 82 66 144 108 135 492
test 3 69 62 147 172 137 498
Lab 4, test 1
test 2
test 3
Lab 7, test 1
test 2
test 3
Lab 8, test 1 72 118 215 304 508
test 2 67 131 200 305 524
test 3 67 123 193 312 571
Lab 9, test 1 53 64 142 122 248 498
test 2 48 61 134 179 225 458
test 3 54 66 138 168 313 543
Lab 10, test 1
test 2
test 3
Lab 11, test 1 122 137 241 376 403 937
test 2 133 226 285 323 513 1290
test 3 210 149 195 408 501 923
Lab 12, test 1 45 49 173 202 371 728
test 2 49 52 181 165 391 692
test 3 56 52 163 237 336 577
Lab 13, test 1
test 2
test 3
Lab 14, test 1 209 208 290 297 437 847
test 2 225 192 257 187 635 902
test 3 280 243 284 274 701 765
Lab 15, test 1
test 2
test 3
Lab 16, test 1
test 2
test 3
Lab 17, test 1
test 2
test 3
Appendix 3
8 (8)

-12 2
Collection of the Original Data y ij Method: M6 (Ds) x10 m /s
Decimal: 2 2 2 2 2 2
Laboratory PC50 PC42 PC35 SF42 FA42 SL42
Lab 1, test 1 1.01 0.92 0.56 0.63 0.24 0.21
test 2 1.1 0.9 0.37 0.71 0.29 0.46
test 3 1 0.96 0.54 0.18 0.28
Lab 2, test 1 1.94 2.73 1.11 1.78 0.77 0.23
test 2 3.35 2.32 1.4 2.4 0.67 0.36
test 3 4.47 4.22 2.02 2.48 0.74 0.36
Lab 3, test 1 0.59 0.41 0.27 0.41 0.19 0.13
test 2 0.65 0.45 0.24 0.28 0.23 0.12
test 3 0.48 0.43 0.26 0.36 0.19 0.15
Lab 4, test 1
test 2
test 3
Lab 7, test 1
test 2
test 3
Lab 8, test 1 1.5 1.62 0.68 0.48 0.25 0.35
test 2 2.2 1.22 1.18 1.18 0.33 0.34
test 3 2.05 1.62 0.62 0.95 0.34 0.32
Lab 9, test 1 0.1 0.88 0.13 0.76 0.08 0.08
test 2 0.1 1.01 0.17 0.62 0.08 0.08
test 3 0.11 1 0.16 0.72 0.09 0.07
Lab 10, test 1
test 2
test 3
Lab 11, test 1 1.24 1.21 2.19 1.96 0.66 0.69
test 2 1.19 1.23 1.57 1.67 0.81 0.47
test 3 1.34 1.28 1.76 2.25 0.81 0.44
Lab 12, test 1 4.58 5.48 1.58 1.87 1.18 0.83
test 2 4.76 4.17 1.95 1.11 0.9 0.8
test 3 4.95 2.88 1.17 1.4 1.48 0.6
Lab 13, test 1 6.19 4.2 1.97 3.02 1.4 0.68
test 2 3.03 4.22 2 1.95 1.39 0.78
test 3 4.25 3.66 1.99 2.36 1.47 0.87
Lab 14, test 1 0.87 1.18 0.28 0.58 0.26 0.17
test 2 1.58 1.2 0.4 1.29 0.14 0.2
test 3 0.67 0.73 0.4 1.03 0.18 0.22
Lab 15, test 1 10.28 2.79 2.72 0.77 0.29
test 2 7.32 3.73 2.38 3.68 0.84 0.31
test 3 9.2 3.1 3.2 3.53 0.73 0.28
Lab 16, test 1 3.37 3.85 3.37 4.17 2.42 0.76
test 2 3.46 4.37 2.96 2.83 1.62 1.03
test 3 5.41 4.1 2.99 2.98 1.46 1.37
Lab 17, test 1
test 2
test 3
Appendix 4
1 (3)
Appendix 4: Received answers to the questionnaire

Questionnaire for the test methods evaluated in WP 5 P1

Please give 1-5 point to each method


D2 M4 M6 R1

Rapidity regarding test duration (most rapid 5 point) 1 4 3 5

Simplicity in equipment requirement and operation (simplest 5 points) 2 4 4 5

Easiness in handling specimens (easest 5 point) 2 4 4 4

Easiness in measurements (easest 5 point) 1 4 3 5


Easiness in calculating results (easest 5 point) 2 4 3 5
Mean 1.6 4 3.4 4.8

Questionnaire for the test methods evaluated in WP 5 P2

Please give 1-5 point to each method


D2 M4 M6 R1

Rapidity regarding test duration (most rapid 5 point) 1 4 4 5

Simplicity in equipment requirement and operation (simplest 5 points) 1 4 4 5

Easiness in handling specimens (easest 5 point) 1 4 5 5

Easiness in measurements (easest 5 point) 1 3 4 5


Easiness in calculating results (easest 5 point) 2 4 4 5
Mean 1.2 3.8 4.2 5

Questionnaire for the test methods evaluated in WP 5 P3

Please give 1-5 point to each method


D2 M4 M6 R1

Rapidity regarding test duration (most rapid 5 point) 2 4 3 5

Simplicity in equipment requirement and operation (simplest 5 points) 3 3 3 5

Easiness in handling specimens (easest 5 point) 2 3 4 5

Easiness in measurements (easest 5 point) 2 4 4 5


Easiness in calculating results (easest 5 point) 3 4 3 5
Mean 2.4 3.6 3.4 5

Questionnaire for the test methods evaluated in WP 5 P7

Please give 1-5 point to each method


D2 M4 M6 R1

Rapidity regarding test duration (most rapid 5 point) 1 4 3 5

Simplicity in equipment requirement and operation (simplest 5 points) 4 1 2 5

Easiness in handling specimens (easest 5 point) 2 1 2 5

Easiness in measurements (easest 5 point) 2 3 4 5


Easiness in calculating results (easest 5 point) 3 4 3 5
Mean 2.4 2.6 2.8 5
Appendix 4
2 (3)

Questionnaire for the test methods evaluated in WP 5 P8

Please give 1-5 point to each method


D2 * M4 M6 R1

Rapidity regarding test duration (most rapid 5 point) 1 4 3 5

Simplicity in equipment requirement and operation (simplest 5 points) 2 5 1 5

Easiness in handling specimens (easest 5 point) 1 3 3 3

Easiness in measurements (easest 5 point) 1 3 2 5


Easiness in calculating results (easest 5 point) 2 4 1 5
Mean 1.4 3.8 2 4.6
Note *: comment on D2 is based on extensive experience in the past
Questionnaire for the test methods evaluated in WP 5 P9

Please give 1-5 point to each method


D2 M4 M6 R1

Rapidity regarding test duration (most rapid 5 point) 1 5 3 5

Simplicity in equipment requirement and operation (simplest 5 points) 1 4 4 2

Easiness in handling specimens (easest 5 point) 1 5 5 3

Easiness in measurements (easest 5 point) 1 5 3 1


Easiness in calculating results (easest 5 point) 1 4 3 5
Mean 1 4.6 3.6 3.2

Questionnaire for the test methods evaluated in WP 5 P11

Please give 1-5 point to each method


D2 M4 M6 R1

Rapidity regarding test duration (most rapid 5 point) 2 4 3 5

Simplicity in equipment requirement and operation (simplest 5 points) 4 3 3 5

Easiness in handling specimens (easest 5 point) 4 3 3 5

Easiness in measurements (easest 5 point) 1 3 3 5


Easiness in calculating results (easest 5 point) 2 4 2 5
Mean 2.6 3.4 2.8 5

Questionnaire for the test methods evaluated in WP 5 P14

Please give 1-5 point to each method


D2 M4 M6 R1

Rapidity regarding test duration (most rapid 5 point) 1 5 3 5

Simplicity in equipment requirement and operation (simplest 5 points) 2 5 2 5

Easiness in handling specimens (easest 5 point) 5 4 4 4

Easiness in measurements (easest 5 point) 3 5 4 5


Easiness in calculating results (easest 5 point) 3 3 3 3
Mean 2.8 4.4 3.2 4.4
Appendix 4
3 (3)

Questionnaire for the test methods evaluated in WP 5 P15

Please give 1-5 point to each method


D2 M4 M6 R1

Rapidity regarding test duration (most rapid 5 point) 1 4 3 5

Simplicity in equipment requirement and operation (simplest 5 points) 1 3 3 3

Easiness in handling specimens (easest 5 point) 2 4 3 5

Easiness in measurements (easest 5 point) 2 3 4 4


Easiness in calculating results (easest 5 point) 2 4 3 5
Mean 1.6 3.6 3.2 4.4

Questionnaire for the test methods evaluated in WP 5 P17

Please give 1-5 point to each method


D2 M4 M6 R1

Rapidity regarding test duration (most rapid 5 point) 3 5 4

Simplicity in equipment requirement and operation (simplest 5 points) 4 5 3

Easiness in handling specimens (easest 5 point) 2 5 4

Easiness in measurements (easest 5 point) 3 5 4


Easiness in calculating results (easest 5 point) 4 5 4
Mean 3.2 5 3.8

You might also like